Home Search Collections Journals About Contact us My IOPscience

Probe measurements of electron-energy distributions in plasmas: what can we measure and

how can we achieve reliable results?

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2011 J. Phys. D: Appl. Phys. 44 233001

(http://iopscience.iop.org/0022-3727/44/23/233001)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 141.225.218.75
The article was downloaded on 11/06/2012 at 07:10

Please note that terms and conditions apply.

IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 44 (2011) 269501 (1pp) doi:10.1088/0022-3727/44/26/269501

Corrigendum
Probe measurements of electron-energy
distributions in plasmas: what can we measure and
how can we achieve reliable results?
V A Godyak and V I Demidov 2011 J. Phys. D: Appl.
Phys. 44 233001
Received 31 May 2011
Published 16 June 2011

On page 4 in equations (4) and (5) as well as in the first line in

the right column, ‘f0 ’ should be replaced by ‘f ’.
On page 4, equation (9) should be replaced by


d 2 Ie 2π e3 Sp e 2 Sp 2e
2
= 2
f0 (eV ) = − F (eV )
dV m 4 mV
e 3 Sp
= √ fp (eV ).
2 2m
On page 5, equation (11) should be replaced by
√ 
4 2m/|e| −∞ 
Te = Ie (V )|V |3/2 dV .
3N Sp 0

0022-3727/11/269501+01$33.00 1 © 2011 IOP Publishing Ltd Printed in the UK & the USA
IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 44 (2011) 233001 (30pp) doi:10.1088/0022-3727/44/23/233001

TOPICAL REVIEW

Probe measurements of electron-energy

distributions in plasmas: what can we
measure and how can we achieve reliable
results?
V A Godyak1 and V I Demidov2
1
RF Plasma Consulting, Brookline, MA 02446, USA
2
Department of Physics, West Virginia University, Morgantown, WV 26506, USA

Received 8 April 2011

Published 16 May 2011
Online at stacks.iop.org/JPhysD/44/233001

Abstract
An electric-probe method for the diagnostics of electron-distribution functions (EDFs) in
plasmas is reviewed with emphasis on receiving reliable results while taking into account
appropriate probe construction, various measurement errors and the limitations of theories.
The starting point is a discussion of the Druyvesteyn method for measurements in weakly
ionized, low-pressure and isotropic plasma. This section includes a description of correct
probe design, the influence of circuit resistance, ion current and plasma oscillations and
probe-surface effects on measurements. At present, the Druyvesteyn method is the most
developed, consistent and routine way to measure the EDF. The following section of the
review describes an extension of the classical EDF measurements into higher pressures,
magnetic fields and anisotropic plasmas. To date, these methods have been used by a very
limited number of researchers. Therefore, their verification has not yet been fully completed,
and their reliable implementation still requires additional research. Nevertheless, the described
methods are complemented by appropriate examples of measurements demonstrating their
potential value.
(Some figures in this article are in colour only in the electronic version)

Nomenclature f electron-distribution function (EDF)

F electron-energy distribution function
a probe radius fjk tensor component of the EDF (fj ≡ fj0 )
A, An instrumental functions ffl typical fluctuation time
b probe-holder radius fp electron-energy probability function
B vector of the magnetic field (|B | = B) fpm measured electron-energy probability function
C filter capacitance f0 , f1 isotropic and directed parts of the EDF (|f1 | = f1 )
Cin input capacitance G gain
Cpr capacitance of probe referenced to plasma IB Bohm current to counter-electrode
Cq coefficient defying class of EEDF Ic displacement current
C0 winding stray capacitance Id discharge current
Da ambipolar diffusion coefficient Ie electron probe current
De diffusion coefficient of electrons (Ie = dIe /dV , Ie = d2 Ie /dV 2 )
e electron charge (−1.6 × 10−19 C) Ieo electron probe saturation current

0022-3727/11/233001+30$33.00 1 © 2011 IOP Publishing Ltd Printed in the UK & the USA
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Ip probe current (Ip = dIp /dV , Ip = d2 Ip /dV 2 ) Vsw voltage range of a sweep
Ir maximal current provided by the Vt voltage of small, time-variable amplitude
reference counter-electrode Xf rf-filter reactance
Iz current corresponding generation of Yjk spherical functions
electrons in plasma volume Zf filter impedance
K number of probe orientations Zpr impedance between probe and plasma
l probe length
L inductance
Lf rf-filter inductance
Greek symbols
m electron mass
M ion mass
γ geometrical factor
MH mass of hydrogen ion
e generation rate of electrons
N plasma density
δ =Rc /Rpo
Nm measured plasma density
ε kinetic energy of electrons
Ns plasma density near chamber wall
ε∗ atom (molecule) excitation energy
N0 plasma density at volume centre
ε average electron energy (ε = (3/2)Te )
Pj Legendre polynomial (Pj ≡ Yj0 )
ϑ polar angle in the spherical coordinate system
Q Q-factor
θ polar angle in equation (24)
r space-coordinate vector
λ angle between axis of cylindrical probe
Rc probe circuit resistance
and symmetry axis of plasma
Rcsh chamber sheath resistance
λe mean free path of electrons
Rext external resistance (includes all resistances
plasma characteristic (diffusion) length
between probe tip and voltage source)
p probe diffusion length
Rf rf-filter resistance
λD electron Debye radius
Rint internal resistance (includes Rcsh , Rx and Rpl )
λε electron-energy relaxation length
RLe electron Larmor radius
λi mean free path of ions
Rpl probe presheath resistance
ν frequency of electron–neutral collisions
Rpo minimal value of probe-sheath resistance
νee frequency of interelectron collisions
Rpsh probe-sheath resistance
τD time scale for electron diffusion to probe
Rt contact resistance
τe electron transient time across probe sheath
Rx contaminated chamber surface resistance
τi ion transient time across probe sheath
Rw resistance of lead wires
τp plasma transient time
Rν sensor resistance
τpN plasma density decay time in afterglow
Sch plasma chamber surface area
τs probe scan time
Sp area of the probe surface
τsw sweep time
t time
ϕ azimuthal angle in a spherical coordinate system
Te electron temperature (in units of energy)
φ plasma potential
Tel low-energy electron temperature
f , c angles defining orientation of probe in plasma
in bi-Maxwellian EEDF
s sink parameter
Tem measured electron temperature
 diffusion parameter
(in units of energy)
ω rf frequency
Tesc electron screening temperature
ωe electron-plasma frequency
(in units of energy)
ωi ion-plasma frequency
Tg gas temperature
ωH electron cyclotron frequency
The high-energy electron temperature
in bi-Maxwellian EEDF
v vector of the electron velocity (| v | = v)
V probe voltage with respect to the plasma Abbreviations
Va probe voltage referenced to ground
vb ion sound (Bohm) speed (vB = (Te /M)1/2 ) dc direct current
Vdc dc voltage rf radio frequency
Vfl voltage of floating probe CCP capacitive coupled plasma
Vext voltage drop across external resistor CW continuous wave
Vint voltage drop across internal resistor ECR electron cyclotron resonance
Vprf rms rf plasma potential referenced to ground EDF electron-distribution function
Vr residual voltage EEDF electron-energy distribution function
Vs plasma-space potential EEPF electron-energy probability function
Vshfr rms rf voltage in probe sheath hf high frequency

2
J. Phys. D: Appl. Phys. 44 (2011) 233001
ICP inductive coupled plasma
MP measurement probe
OD outer diameter
PVS probe-voltage source
RVS residual voltage source
RP reference probe
1. Introduction
An electric probe is essentially a conducting object immersed
into or facing (as in the case of a wall probe) plasma
for the purpose of diagnostics. The first, not particularly
successful attempts to use floating electric conductors for
the measurement of plasma-space potentials Vs were made
at the beginning of the 20th century (see, e.g., [1]), but
the first real progress in obtaining plasma parameters from
probe measurements was made later by Langmuir et al [2, 3].
Therefore, these probes are commonly referred to as Langmuir
probes.
Langmuir demonstrated that, for spherical and cylindrical
probes inserted into a weakly ionized low-pressure plasma in
which the dimension of the plasma volume distorted by the
probe is significantly smaller than the mean free path of the
electrons λe (i.e. electrons from the undisturbed plasma can
reach the probe surface without collisions), the application of
a voltage V to the probe and the measurement of the probe
current Ip can provide reliable information about not only Vs
but also the plasma density N and the electron temperature Te .
The theory developed by Langmuir led to the next important
step in the development of probe diagnostics, which was
undertaken by Druyvesteyn [4], who demonstrated that the
second derivative of the probe current with respect to the probe
potentials d2 Ip /dV 2 ≡ Ip allows the determination of the
electron-energy distribution function (EEDF) in the plasma.
The electric probe was seen as a rather simple and
attractive scientific instrument, and after seminal works by
Langmuir and Druyvesteyn, numerous probe measurements in
gas-discharge plasmas have been conducted by many authors.
Today, the field of electric probes is enormous, and many
works on the subject can be found in the literature (see,
e.g., the reviews and books [5–27] and references therein).
Druyvesteyn method has been used for measurements of
EEDFs in dc plasma of positive column in noble and molecular
gases [5], including striations [28]; plasma near anode [29]
and cathode [30]; hollow cathode plasma [31]; afterglow [32];
photoplasma [33]; plasma with negative ions [34, 35] and in
many other papers (see above reviews and monographs). Since
the time of Langmuir and Druyvesteyn, probe theory has been
extended dramatically, allowing the diagnostics of different
types of plasmas, including those experiencing high pressure,
strong magnetic fields and anisotropy. More sophisticated
probe constructions (some of which can hardly be considered
as Langmuir probes) allow measurements in chemically active
plasmas (e.g. by a heated probe [12, 24, 25, 36]) and under
the harsh condition of fusion-related plasmas (reciprocating
probe [37]). Currently, electric probes can measure not only
the basic plasma parameters, such as N , Te , Vs and EEDF, but,
depending on the plasma type, can also measure other fluid
3
Topical Review
observables, such as the flow velocities of ions (Mach probe
[20, 25]), ion temperatures (ion-sensitive probe [25, 38]) and
oscillations of N , Te , Vs in magnetized plasma (magnetically
insulated baffled (MIB) probes [39] such as Katsumata probes
[38, 40], plug probes [41, 42], baffled probes [43, 44], tunnel
probes [45] and ball-pin probes [46]). The potential of the
electric-probe method has not been yet fully exploited and has
great possibilities for development.
Despite the development of new probe constructions and
theories, which are useful for many specific applications and
branches of plasma physics, simple cylindrical Langmuir
probes are currently the main contact diagnostic tool for
measuring plasma parameters in weakly ionized, low-pressure
plasmas in both applied and basic plasma research. For non-
equilibrium plasmas with non-Maxwellian electron-energy
distributions, the measurement of the EEDF is a unique
way to obtain the basic plasma parameters and the rates of
plasma-chemical processes without assuming a Maxwellian
EEDF, instead using the corresponding integrals of the EEDFs
measured experimentally.
However, the interpretation of the measurements for even
this simple case can be intricate and confusing. Therefore,
it is not surprisingly that erroneous results and incorrect
applications of Langmuir probes are common in the literature.
The errors mainly arise from the poor design of probe
experiments, incorrect probe constructions and circuit designs
and a lack of awareness of possible error sources. Sometimes,
the applicability and limitations of probe theories used for
the evaluation of plasma parameters are ignored. Although
these issues have been considered in many probe reviews
and in specific review of EEDF measurements [19], many
recent papers on EEDF measurements leave these problems
unaddressed. At the same time, for the last twenty years
after the publication of review [19], there has been significant
progress in probe experimental design, effectively addressing
the aforementioned problems mainly in relation to rf plasmas
typical for plasma-processing reactors.
The purpose of the present review is to analyse the
most common sources of error in EEDF measurements
and possible ways to avoid them by proper design of
the probe experiment. Additionally, we consider recent
developments of EEDF measurements in complicated plasma
conditions. Among them are measurements in time-variable,
collisional, magnetized and anisotropic plasmas and electron-
spectroscopy analysis. We also present numerous examples of
EEDF measurements obtained with laboratory and commercial
probe instruments in which possible sources of errors were
adequately addressed. In many aspects, the present review
summarizes the authors’ experiences gained over decades of
practicing and developing probe diagnostics.
The paper is structured as follows. In the next section,
we briefly describe the most detailed information that can
be obtained from the probe measurements, namely the
electron-distribution function, EDF and its derivatives, such
as the EEDF. We also provide the primary formula for the
Druyvesteyn method suitable for measurements in weakly
ionized, low-pressure and isotropic plasma. Section 3 deals
with the main sources of error in the probe measurements
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

and their mitigation. Section 4 demonstrates principles
of the probe-diagnostics method beyond the limitations of
the Druyvesteyn procedure, such as EEDF measurement in
collisional, isotropic or magnetized plasmas and electron-
spectroscopy diagnostics. The last chapter of this work
provides concluding remarks.
2. General considerations
2.1. What we are going to measure
Electron gas in plasma is described by the electron-distribution
function (EDF) f (t, r, v), where r is the space-coordinate
vector and v is the velocity vector [47, 50]. The function
gives the number of electrons in the volume element d3 r =
dx × dy × dz having velocities between v and v + d v at
time t. This function can be used for the description of
all kinetic processes with electron participation, including
electron transport and the excitation of atoms and molecules in
both isotropic and anisotropic plasmas. The electron density
can be found from the function as


N(t, r) = f (t, r, v) d
v. (1)
v
In practice, it is convenient to express the EDF as a series [47],
∞ 
 j
f (t, r, v) = fjk (t, r, v)Yjk (ϑ, ϕ),
j =0 k=−j
where v = | v | is the electron speed, fjk are tensor components
of the EDF, Yjk are spherical functions (harmonics), and ϑ and ϕ
are polar and azimuthal angles in a spherical coordinate system.
In many cases, due to electron collisions with molecules and
plasma boundaries, the EDF is nearly symmetrical and can be
represented by the two-term approximation
In this case, functions f0 and F have identical information
about the properties of the electron gas because the angular
distribution for electrons is the same in any direction.
Consequently, in the isotropic plasma, the electron density can
be found as the following (omitting coordinates and time):

∞ √
∞√
−3/2
N= F (ε) dε = 4m 2π εf0 (ε) dε. (5)
0 0
√ √
The function 4m−3/2 2πf0 (ε) = F (ε)/ ε ≡ fp (ε) is
frequently referred to as the electron-energy probability
function (EEPF) [19, 48–50]. For isotropic plasma, it has the
same information about the electron gas as EDF or EEDF and is
frequently used to represent measured probe data. The EEPF,
presented in a semi-log scale, allows the quick visualization
of a departure of the measured EEPF from a Maxwellian
distribution, which is a straight line in this representation. Note
that some authors use different normalizations of the above
distribution functions, but we believe that the representation
given above is the most convenient way to read numerous
published probe records. Knowledge of the EEDF or EEPF
allows the calculation of the plasma parameters and the rates
of plasma-chemical processes. Thus, for plasma density N and
effective electron temperature Te (defined for non-Maxwellian
EEDF as a measure of the average electron energy ε as
Te = (2/3)ε), the following expressions hold:

∞
N= (ε)fp (ε) dε, (6)
(2) 0
and


∞
2 −1 ∞ 3/2 2
Te = N ε fp (ε) dε = N −1 εF (ε) dε. (7)
3 0 3 0
Similarly, the collision frequencies for electron–atom elastic
collisions, excitation and ionization frequencies and the
electron screening temperature Tesc can be calculated as
appropriate integrals of the measured EEPF [51].

v  2.2. The Druyvesteyn method

f (t, r, v) = f0 (t, r, v) + · f1 (t, r, v). (3)
v
Probe measurement of EEDF/EEPF in weakly ionized, low-
This representation is valid even for rather low gas pressures pressure, isotropic plasma is based on the Langmuir expression
because most electrons are imprisoned in the plasma for electron probe current [2, 3],
volume due to the near-wall sheath. Therefore, the two-

2π eSp ∞
term approximation is generally accepted in most plasma Ie = (ε − eV )f0 (ε) dε
m2
research. In some special cases (e.g. for very low pressures,
eV∞
eSp F (ε)
strong electric fields and electron beams), the two-term = √ (ε − eV ) √ dε
approximation is not enough and the above expansion to 2 2m eV ε

∞
spherical harmonics (general case) or to Legendre polynomials eSp
= √ (ε − eV )fp (ε) dε. (8)
Pj for cylindrical symmetry (in this case, k = 0 and Pj ≡ Yj0 ) 2 2m eV
should be used. Here, Sp is the probe area. It is assumed that the probe
In line with the EDF, the use of the EEDF, F (t, r, ε), is negatively biased (V < 0), and e is the electron charge
which gives a number of electrons in the volume element (−1.6 × 10−19 C). The probe current is assumed to be directed
having energies between ε and ε +dε (losing information about to the probe. Double differentiation of equation (8) on the
angle distribution), is generally accepted. For an isotropic probe potential V gives the Druyvesteyn formula [4]:
EDF,√f (t, r, v) = f0 (t, r, v) and, taking into account that 
d 2 Ie 2π e3 Sp e2 Sp 2e
v = 2ε/m, where ε is the electron kinetic energy and m is = f0 (eV ) = − F (ε)
dV 2 m2 4 mV
the electron mass, the following expression holds: 
√ √  e2 Sp ε 2e
=− fp (ε). (9)
F (t, r, ε) = 4m−3/2 2π εf0 (t, r, 2ε/m). (4) 4 mV

4
J. Phys. D: Appl. Phys. 44 (2011) 233001
Thus, measurements of Ie can provide
√ noise are practically invisible on the measured probe I /V


2 2m −∞ 
N= Ie (V ) V /e dV (10)
|e|Sp 0
and √

4 2m −∞ 
Te = Ie (V )(V /e)3/2 dV . (11)
3N Sp 0
There is another method of inferring EEDF directly from
equation (8) by solving this integral equation using different
regularization (deconvolution) procedures. The reader can
find more on the subject in [21, 52]. To date, this method
has not been routinely used in EEDF measurements due to
its inferiority to EEDF measurement using the Druyvesteyn
method.
2.3. What makes a good EEDF measurement?
EEDF measurements yield meaningful results only when they
contain accurate information about the majority of electrons
in both elastic (ε < ε∗ ) and inelastic (ε > ε ∗ ) energy ranges,
where ε∗ is the excitation energy. The EEDF in the energy
range between ε = 0 and ε = Te contains the majority of
electrons and is responsible for evaluation of electron density
and transport process. The distortion of this part of EEDF,
which can be dramatic, significantly affects the accuracy of
corresponding data found from the measured EEDF. The high-
energy tail of the EEDF defines inelastic processes (such as
excitation and ionization) and is very important for the kinetics
of the excited and ionized states of atoms and molecules.
Acceptable EEPF data require that the energy gap between
a zero point and the peak of the second derivative of the probe
current not exceed (0.3–0.5)Te , and that the high-energy tail
beyond the inelastic threshold is not masked by noise and/or by
ion current. This requirement calls for an EEPF-measurement
instrument with a high dynamic range of approximately
60–80 dB and high-energy resolution of a fraction of Te .
Most of the published EEPF data obtained from laboratory
experiments and commercial reactors (using homemade and
commercial probe instruments) lack information about the
bulk of low-energy electrons and the high-energy tail. The
calculation of plasma parameters from these distorted EEPFs
leads to errors and even the inability to analyse the EEPF tail
responsible for excitation and ionization processes.
The simplicity of the concept upon which probe
diagnostics are based has promoted the widespread illusion
that the measurement, processing and interpretation of probe
characteristics are commonplace and routine. Indeed, there
is no plasma diagnostic method other than probe diagnostics
where the danger of incorrect measurements and erroneous
interpretation of the results is so great [8]. This statement,
even in to a greater extent, is true in the measurement of
EEDF due to the effects of error augmentation inherent to
differentiation and regularization (deconvolution) procedures.
In the measurement of probe I /V characteristics, it is
important to realize that even small errors, which are tolerable
in classic Langmuir-probe diagnostics, can result in enormous
distortion in the inferred EEDF. Deterioration effects of probe
5
Topical Review
contamination, circuit resistance and low-frequency and rf
characteristic but are usually clearly seen in the resultant EEDF.
Therefore, to obtain a reliable EEPF, special attention must be
paid to the accuracy of the probe measurements themselves
and to the correct application and limitations of the traditional
(Langmuir) probe method. Therefore, the next section is
devoted to the consideration of the most common problems
in probe measurements of EEDF and the methods of their
mitigation.
3. Problems in probe measurements and their
mitigation
3.1. Probe size
The inference of the EEDF from the measured volt/ampere
characteristic of the probe immersed in plasma implies the
validity of the very same assumptions as for the classical
Langmuir-probe diagnostics. These assumptions require
that the probe is small enough to avoid plasma and sheath
perturbations beyond those accounted for in the probe theory
[53, 54] (the application of probes for higher pressure and more
exact evaluation of possible errors are discussed in section 4).
For a typical cylindrical probe, the following inequalities
must be satisfied for the fulfilment of the small-probe
assumption:
 
πl
a ln , b, λD λe and Ip Id , Ir , Iz . (12)
4a
Here, a is the probe-tip radius, l is its length, b is the probe-
holder radius, λD is the electron Debye radius, Ip is the
probe current, Id is the discharge current, Ir is the maximal
current provided by the reference counter-electrode in the
probe circuit (ion current to the grounded metal chamber,
Ir = IB = Sch eNs vB , or electron-emission ability of the
auxiliary cathode), and Iz = ee is the current corresponding
to the generation rate of electrons e with energy ε in the
volume defined by the chamber characteristic size or by the
electron relaxation length λε , whichever is shorter [19, 25, 55].
Here, Sch is the chamber surface area, Ns is the plasma density
near the chamber wall, vB is the ion sound (or Bohm) speed
(vB = (Te /M)1/2 ), M is the ion mass and IB is Bohm current
on counter-electrode. The value of e is sensitive to the
electron energy but must be equal to IB /e after being averaged
over EEPF. For small discharge chambers, the flux of the
fast electrons drawn to the probe may be comparable to their
generation rate due to heating by the electromagnetic field
and electron–electron collisions. In this case, the measured
EEPF may be depleted with the high-energy electron tail. Note
also that conditions (12) may be significantly more severe in
electronegative gases [56]. In this case, negative ions can
be repelled from the probe-holder at a distance significantly
greater than λe , and the charged-particle density can be
severely distorted near the probe surface, even if inequality
(12) holds.
With increasing gas pressure and when the ratio of the
plasma density at its boundary to that at the plasma centre,
Ns /N0 1, decreases, the wall-sheath resistance may not
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

be negligible compared with the probe-sheath resistance Rpsh

in the plasma centre. This situation leads to a redistribution
of the voltage applied to the probe between the probe and
the wall sheaths and results in a reduction of the electron
current and a corresponding rounding of the measurement of
the second derivative of the probe current in the vicinity of
the plasma-space potential. The maximal sheath-resistance
ratio at the chamber wall and the probe, Rcsh /Rpsh = Ip /IB ,
occurs when the probe is at the plasma potential at which
Rcsh /Rpsh = (Sp N0 /Sch Ns )(M/2πm)1/2 , where Sp = 2π al.
Thus, to neglect the voltage drop across the wall sheath, the
following requirement must be satisfied:

(Sp N0 /Sch Ns )(M/2πm)1/2 1. (13)

Note that inequality (13) is derived for Maxwellian EEPF. In

argon (and in other Ramsauer gases), a CCP at low pressure has
a bi-Maxwellian EEPF with a low-energy peak corresponding
to the electron temperature Tel and a tail of hot electrons
corresponding to electron temperatures The , where Tel The .
In this condition, the probe resistance at the plasma potential Figure 1. Telescopic probe P1 with a ring reference probe P2 used
is defined by Tel , whereas the sheath resistance at the wall with for EEDF measurement in CCP [55].
the floating potential reference to plasma is defined by Teh .
Here, inequality (13) becomes even stronger; (Tel /The )1/2 < 1
should be on the right-hand side instead of 1. The effects of
a final probe circuit resistance associated with the violation of
inequality (13) on the distortion of the measured EEPF and a
remedy to relax inequality (13) will be discussed later in this
review.
The first inequality in (12) means that there are no
electron–atom collisions in the probe perturbation area Figure 2. Probe with a thin-wall quartz protective sleeve used in the
EEDF measurement in CCP [55].
including the probe sheath and presheath. Note that this
inequality is applied not only to the probe tip that collects
the probe current but also to an adjacent insulated probe
holder. In many published works, the probe-holder diameter
is excessively large, causing plasma perturbation near the
collecting probe tip. This perturbation is similar to that
occurring near the chamber wall (the probe holder is acting
as an undesirable wall) and occurs due to the plasma sink to
the probe holder.
Examples of a probe design with minimized plasma Figure 3. Telescopic probe used for EEDF measurement in argon
perturbation near the probe are depicted in figures 1 and 2, ICP [55].
which show the telescopic probe structure with a thin probe
holder used for EEDF measurements in symmetrical CCP with
argon gas [55]. Here, the probe holder is made of a thin-wall
quartz capillary with a diameter of b = 0.33 mm and a length of
2 cm that transits to a stronger structure with a larger diameter.
Figure 4. Low-perturbation rf-compensated probe for measurement
The probe has a sleeve that prevents electrical contact between in plasma reactors [60]. P1 is the measuring probe with
the probe and the sputtered conductive material on the probe- a = 0.05 mm. The probe holder has radius of b = 0.5 mm. P2 is the
holder surface. This sleeve is a necessity when measuring reference probe for low-frequency noise suppression and acts as a
in CCP due to considerable electrode sputtering by ions shunting electrode for rf compensation.
accelerated in the electrode rf sheath. The probe structure
includes the measurement probe P1 and a ring reference probe basic ICP studies [51] and that developed for probe diagnostics
(RP) P2 used for low-frequency and rf noise suppression in the in plasma-chemical rf reactors [60] are shown in figures 3 and 4,
probe-measurement circuit that will be discussed later. respectively.
The probe diameter 2a and length are 76 µm and In the telescopic probe structure used for measurement
6 mm, respectively. The results of the EEDF measurements in argon ICP [51] shown in figure 3, the probe holder near
performed with this probe can be found in [51, 55, 57–59], and the probe is twice as thin as that used in CCP. Despite the
some will be shown later. The low-perturbation probe used for small probe tip (a = 38 µm, l = 6 mm, chamber conductive

6
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 5. Probe-measuring configuration in plasma within a metal
chamber (left) and its equivalent circuit (right) [51]. RVS is a
residual voltage source, PVS is probe-voltage source, Rpsh is
probe-sheath resistance, Rext is external resistance (includes all
resistances between probe tip and voltage source) and Rint is internal
resistance.
surface Sch ≈ 0.1 m2 ) and the electronics compensating the
voltage drop across the chamber-wall sheath, the upper limit
of argon pressure at which measurement of EEDF is possible is
300 mTorr. At higher gas pressure, the electron current to the
probe that is limited by the ion saturation current to the chamber
wall prevents the probe from reaching the plasma potential.
The volt–ampere characteristic of the wall sheath is that of
an ideal flat probe corresponding to near-infinite differential
resistance in the area of the ion saturation current. For this
reason, it is impossible to reach the plasma potential even after
applying a very high voltage on the probe; the only result is an
increase in the voltage across the chamber-wall sheath.
3.2. Probe circuit resistance
The resistance of the sheath at the chamber wall (or next to any
other counter-electrode in the probe-current path) is only one of
many causes of distortion in the measured probe characteristic
and its second derivative. When making probe measurements,
one usually assumes that the voltage applied to the probe is
localized in the probe sheath around the probe tip. In reality,
this voltage is distributed along different parts of the probe-
current path. As a rule, the probe current and voltage measured
directly at some point of the probe circuit are not the same as
the current and voltage across the probe sheath.
The probe circuit configuration in application to an ICP
with a grounded chamber and its equivalent circuit diagram
are shown in figure 5. The probe current Ip flows from the
grounded probe-voltage source (PVS) back to ground through
various circuit elements along the current path: the external
resistance Rext (which includes all resistances located between
the probe tip and the PVS), the probe-sheath resistance Rpsh
and the internal resistance Rint consisting of the chamber
sheath resistance Rcsh , the contaminated chamber surface
resistance Rx and the probe presheath Rpl resistance, Rint =
Rcsh + Rx + Rpl . The plasma potential source, RVS (generally
containing dc, low-frequency and rf components), is shown
in the equivalent circuit as a residual (noise) voltage source.
7
Topical Review
The external resistance is Rext = Rw + Rt + Rv + Rf , where
Rw is the resistance of the lead wires, Rt is the contact
resistance, Rv is the sensor resistance and Rf is the rf-filter
resistance at dc and low frequencies. Note that the rf-filter
reactance Xf = (Lf /Ip )dIp /dt may also affect the probe circuit
impedance at fast probe-voltage scans.
As a result of the aforementioned resistances, except for
the probe-sheath resistance, the total probe circuit resistance
Rc = Rext + Rint = Rw + Rt + Rv + Rf + Rcsh + Rx + Rpl .
Some of these resistances (such as Rw , Rt , Rv and Rf ) are
constant during the probe scan and can be accounted for by
processing the probe characteristic. Others (such as Rcsh
and Rx ) are unknown, sometimes non-linear (thus, probe-
current dependent) and generally not easy or even impossible
to evaluate and account for. In practice, the probe voltage
referenced to the ground, Va , is measured at the far end of the
probe lead wire, point ‘a’ in figure 5. Therefore, it includes
not only the voltage drop across the probe sheath V but also
the voltage drops across the external resistor Vext = Ip Rext ,
the voltage across the internal resistance Vint = Ip Rint and the
residual voltage Vr . Thus, the voltage applied to the probe is
Va = V + Vext + Vint + Vr . To infer the true voltage across
the probe sheath V , one has to account for stray voltages
Vext , Vint and Vr , which are not always negligible. On the
other hand, the presheath plasma resistance Rpl is negligibly
small in low-pressure gas-discharge plasmas, Rpl /Rpo ≈
πl
(a/λe ) ln( 4a ) 1, where Rpo is the minimal value of the
probe-sheath resistance.
The probe circuit resistance significantly affects the probe
I /V characteristic near the plasma potential where the probe
current reaches its maximum and the probe-sheath resistance
its minimum. For Maxwellian plasma, the minimal probe-
sheath resistance Rpo = dV /dIp |V =0 = Te /eIpo , where
Ipo = Ip |V =0 is the electron saturation current. Due to error
magnification inherent to the differentiation procedure, even
a relatively small distortion in the I /V probe characteristic
results in an enormous distortion in the inferred EEPF. This
distortion manifests itself in the suppression and flattening
of the second derivative of the probe I /V characteristic,
d2 Ip /dV 2 , near the plasma potential and an increased
voltage interval between the second-derivative maximum
and the zero-crossing point. The EEPF is proportional to
d2 Ip /dV 2 and, when distorted in such manner, looks like a
Druyvesteyn-like distribution (EEPF ∼ exp(−ε 2 /ε2 )). This
‘Druyvesteynization’ of an EEPF with increasing discharge
power and/or gas pressure can be seen in the measurements of
many authors and has even been mistaken for a new plasma
kinetic effect in ICP.
EEPF depletion due to probe circuit resistance is
proportional to (V )3 , where V is the voltage drop across
the circuit resistance and is most significant near the plasma
potential, which corresponds to the low-energy electrons
of the distribution. EEPF distortion depends on the ratio
Rc /Rpo , where Rpo is the minimal probe differential resistance
(at plasma potential) Rpo = Te /eIeo and Ieo is the electron
saturation current. The distortion of d2 Ip /dV 2 for Maxwellian
EEDF caused by final probe circuit resistance is demonstrated
in figure 6. Here, the distorted shapes of d2 Ip /dV 2 that are
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 6. Second derivative (absolute value) of the probe electron
current for Maxwellian EEDF for different parameters δ = Rpo /Rc .
The Druyvesteynization effect is seen for large δ [55].
proportional to EEPFs are shown for the different parameters
δ = Rc /Rpo . The EEPFs are dramatically distorted when
the probe circuit resistance approaches the probe-sheath
resistance. A calculation shows that the depletion of the second
derivative of the probe current at the plasma potential due to
probe circuit resistance by less than 3% requires the total probe
circuit resistance Rc to be one hundred times smaller than Rpo .
This finding implies that, for undistorted EEDF measurements
in high-density plasma, typical of ICP and Helicon plasma
sources, Rc should be in the range of a few milliohms to tenths
of an ohm. This requirement may not be easily achieved for
a probe circuit having many resistive components, such as a
filter choke and plasma to chamber-wall resistance.
This problem is severe in ICP with plasma-processing
gases, where the metal chamber wall is covered with a low-
conductivity layer of the plasma reaction products (large
Rx ). Even in argon ICP with clean chamber walls, the wall-
sheath resistance Rcsh is frequently comparable to the probe-
sheath resistance near the plasma potential, resulting in the
Druyvesteynization of the measured EEPF. The wall-sheath
resistances, Rcsh and Rx , also limit the maximum accessible
gas pressure for ICP probe diagnostics when the ion current to
the chamber wall is less than the electron saturation current to
the probe, considered above.
The Druyvesteynization of EEPF and saturation of the
plasma density with increasing gas pressure found in some
published ICP probe measurements is generally the result
of the finite probe circuit resistance. This effect is more
pronounced at relatively high gas pressures when the plasma
density near the chamber wall, Ns , is much smaller than that in
the plasma bulk (Ns N0 ). Note that, in bounded diffusion-
controlled plasmas, Ns /N0 ∼ λi / when λi / < Tg /Te and
Ns /N0 ∼ (λi / )1/2 when λi / > Tg /Te , where N0 is the
plasma density in the discharge centre, λi is the ion mean free
path and Tg is the gas temperature [61].
The problem of the unaccounted resistance of the probe-
current path remains unresolved in majority of ICP (and
in other high-density plasma sources) probe experiments
performed with homemade and commercial probe systems,
8
Topical Review
Figure 7. Probe driver circuit with a differential input for
compensation of the dc plasma potential, low-frequency noise and
internal resistance. The reference probe feeds into the positive input
while the measuring probe voltage feeds into the negative input,
providing the driver output with the difference between the ramp
and plasma potential voltages [51] . MP is a measurement probe, RP
is a reference probe and Rv is a sensor resistor.
resulting in a large EEPF distortion in its low-energy range.
This distortion manifests itself as an enormous depletion (and
even, absence) of low-energy electrons in the measured EEPF,
whereas low-energy electrons constitute the overwhelming
majority of the distribution. Consequently, using large
Langmuir probes exacerbates the problems associated with the
circuit resistance.
Different approaches (for different elements in the probe-
current path) are employed in the probe experiment to address
the issue of circuit resistance. The residual voltage Vr (that
consists of dc, low frequency and rf voltages developed by
RVS) and the voltage drop Vint across the internal resistance
define the instantaneous value of the plasma potential reference
to ground Vpl = Vr + Vint , can be cancelled by a probe-
voltage driver having differential input, as shown in figure 7.
The negative input of the driver is fed with the source of
the PVS, whereas the RP signal corresponding to its floating
potential (point c in figure 5) is fed to the positive input, thus
producing a difference between these two input signals at the
driver output. This difference cancels the dc voltage bias,
low-frequency noise and the chamber-wall voltage drop. For
effective noise cancellation the driver should have sufficient
bandwidth in frequency and phase space. The dc bias and
the chamber-wall voltage-drop compensation are limited by
the maximum output voltage available from the driver. The
probe driver shown in figure 7 does not compensate for
time variation in the probe current due to plasma-density and
electron-temperature (or EEDF) oscillations. Therefore, it can
be used for the diagnostics of periodically varied plasmas in
a time-resolved mode with appropriate signal-acquisition and
processing probe waveforms [62]. On the other hand, this
measurement technique has limitations when working with
strongly unstable plasma. It is impractical to use the RP
feedback for the purpose of rf-noise suppression associated
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Figure 9. Plasma density and effective electron temperature

Te = 2/3ε found as the corresponding integrals of the measured
EEPFs shown in figure 8.

Figure 8. EEPFs measured in argon ICP between 0.3 and

300 mTorr. ε ∗ and εi denote the excitation and ionization energies,
respectively. Vertical lines show the dc plasma potentials referenced
to a grounded chamber [51].

with the discharge driving frequency and its harmonics due to

the speed limitation of the available electronics. Approaches to
mitigating the effects of internal resistance and low-frequency
noise, considered above, have been implemented in past EEDF
measurements in CCP [55], in ICP [51] and in commercial
instrument VGPS of Plasma Sensors® [60].
EEPFs measured in argon ICP driven at 6.78 MHz at
discharge power of 50 W are shown in figure 8 for wide range
of gas pressures [51]. The measurements were performed
with the probe shown in figure 3 and with compensated
internal probe circuit resistance, in a Faraday-shielded ICP
with negligibly low rf plasma potential and no need for probe
rf compensation. The basic plasma parameters obtained from
the measured EEPFs, N and Te , are shown in figure 9.
Note the presence of a well reproduced Maxwellian
distribution in the low-energy part of the EEPFs at relatively
high argon pressure, where the plasma density is sufficiently
large to provide a strong Maxwellizing effect due to electron–
electron (e–e) collisions. The frequency of e–e collisions νee
is proportional to N ε −3/2 ; therefore, for electron energies of
less than average one ε and for N (ε/e)−3/2 larger than
approximately 1010 (cm V2 )−3 , the EEPF must be Maxwellian.
The larger the plasma density, the larger is the energy range
where the EEPF coincides with a Maxwellian distribution.
Figure 10(a) shows another example of EEPFs measured Figure 10. EEPFs measured with different probe systems in an
in an industrial ICP reactor with the VGPS of Plasma Sensors® industrial ICP etcher for a range of discharge powers between 0.25
probe instrument [63] with an rf-compensated probe, shown and 2 kW. The dashed lines show the noise threshold that defines the
maximal resolved electron energy εmax . Note a Maxwellization
in figure 3, and compensated Rint . For comparison, the result effect with increasing rf power at high electron energies
of the measurement made in the same plasma with the Espion (in figure 10(a)), in contrast to the Druyvesteynization effect at low
of Hiden® probe system is shown in figure 10(b). A dramatic electron energies (in figure 10(b)).

9
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 11. EEPFs measured in argon ICP in a wide range of
discharge powers [44]. The EEPFs are shown in doubling power
increments. Note the EEPF Maxwellization in the elastic energy
range (ε < ε∗ ) with increasing rf power [51].
distortion in the low-energy portion of the EEPF, showing a
Druyvesteyn-like distribution in figure 10(b), is apparent. The
difference between the measurements in figures 10(a) and (b)
and the divergence between the measured and Maxwellian
distribution in figure 10(b) increase with rf power and, thus,
with plasma density. The finding in figure 10(b) is obviously
opposite to the well-known trend of EEDF behaviour in gas-
discharge plasmas. The Maxwellizing effect, increasing with
plasma density, is clearly seen in the EEPF measured in a
wide range of discharge powers in argon ICP presented in
figure 11 [51].
The RP does not compensate for the voltage drop in
external resistance Rext . Fortunately, however, this external
resistance can be readily measured and accounted for in probe
characteristic processing. It can also be accomplished with
an analogue circuit known as a gyrator, which is a current-
to-voltage converter with a negative input resistance equal to
Rext [55].
3.3. Plasma oscillation
Probe characteristic distortion in rf plasma is a well-
recognized problem and has been analysed in many works
[19, 21, 55, 64–69]. Nonetheless, this issue has not been
properly addressed in many published experiments performed
with rf plasmas. Some authors believe that simply putting
some filter tuned on the discharge driving frequency and its
harmonics into the probe circuit is sufficient to address the
probe distortion associated with rf oscillation. Indeed, it has
been shown experimentally [65] that, to avoid a noticeable
distortion in EEDF measurement, the rms rf voltage in
the probe sheath, Vshrf , should be two to three times less
10
Topical Review
than the electron temperature expressed in volts, Vshrf <
(0.3–0.5)Te /e. To fulfil this requirement, the probe filter
system must satisfy to the following relationship between the
filter and probe parameters:
|Zpr /Zf |
(0.3 − 0.5)Te /| eVprf |. (14)
Here, Zpr is the impedance between the probe and plasma,
Zf is the filter impedance, Vprf is the rms rf plasma potential
referenced to ground.
There are several ways to minimize Vshrf . One is to design
a rf plasma system with a reduced or even negligibly small rf
plasma potential Vprf . Examples of this approach have been
demonstrated in [51, 68] and partially in [55]. The second
and most common approach is the use of rf filters mentioned
above [19, 55, 66, 69]. Another method is the biasing of the
measuring probe with rf voltages (of fundamental frequency
and few upper harmonics) with amplitudes and phases equal
to those of the plasma rf potential [70, 71]. This approach
is preferable for measurements in rf plasma with a large
rf-electric-field gradient, such as near the electrode area of
a CCP and the skin layer of an ICP, where a large reference
shunting electrode could introduce a considerable disturbance
to the narrow localized rf field.
Now, let us consider the method of the realization of
inequality (14) for a passive filter design. Relation (14) is
practically impossible to satisfy when Zpr is defined solely
by the probe-sheath impedance, usually corresponding to a
reactance of the probe-sheath capacitance of less than 1 pF.
Therefore, a shunting electrode is placed in the probe vicinity
or in the area that is rf equipotential with the rf plasma
potential near the measuring probe [19, 21, 55, 69]. The
shunting electrode is connected to the probe with a capacitor
whose capacitance is much larger than the probe-sheath
capacitance but is small enough not to introduce a noticeable
stray displacement current into the probe-measurement circuit.
With the shunting electrode, the impedance of the probe
referenced to plasma can be reduced by one to two orders of
magnitude.
According to (14), the required filter impedance is defined
by the value of Vprf for each harmonic of the plasma potential
spectrum to be measured. Unfortunately, the straightforward
way of measuring of Vprf (by connecting a scope rf probe to the
measuring Langmuir probe) usually results in a value that is an
order of magnitude lower than the true rf plasma potential. The
reason for this error is that the input capacitance of a rf probe
Cin is much larger than the capacitance of the Langmuir probe
referenced to plasma Cpr , whereas the reverse relationship is a
must for the correct measurement of Vprf . With typical Cpr ≈
1 pF and Cin ≈ 15 pF (for a 1 : 10 rf probe), Vprf measured this
way is approximately 16 times less than its true value.
The authors recommend the following procedure to
measure the true rf plasma potential (see also [72]).
Temporarily insert into the plasma a relatively large probe
(electrode), perhaps 3–6 mm OD, and a 5–10 cm long metal
road (tube) connected via a 1 : 100 rf voltage probe-divider
(usually having Cin = 3–5 pF) to the high-impedance input
of a spectrum analyser. Depending on the electrode size and
plasma density, its capacitance Cpr will be on the order of
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 12. Plasma potential spectrum measured in argon CCP
driven at 13.56 MHz [55].
100 pF, providing 5% accuracy in the measurement of Vprf (for
each harmonic) with a voltage divider having Cin = 5 pF.
The probe–plasma coupling impedance at the probe
floating dc potential is practically defined by the shunting
electrode impedance |Zpr | = [(ω Cpr )2 + (eIB /Te )2 ]−1/2 and
can be readily evaluated for a known electrode surface and
expected plasma parameters N and Te (see, e.g., [25]). The
second term in the expression for Zpr is due to the electrode-
sheath conductivity and dominates at frequencies lower than
the ion-plasma frequency, ω < ωi . With values of Zpr and
Vprf , one can determine the required minimal filter impedance
for each harmonic of Vprf whose rms value is larger than
(0.3–0.5)Te /e
Zf  (2 − 3)Zpr | eVprf |/Te . (15)
To maximize the filter impedance, parallel resonance tanks
tuned to each relevant harmonic are used: Zf = QωL =
Q(ωC)−1 . Here, L is the inductance, C is the filter capacitance
and Q is its Q-factor. Subminiature inductive chokes with self-
resonance frequencies matching the plasma potential spectrum
are the most suitable for filter construction. Their winding stray
capacitance C0 is around 1 pF, and Q is around 10. Having a
minimal possible resonating capacitance and a relatively large
bandwidth ω = ω/Q, they provide high filter impedance
even considering some detuning caused by the filter enclosure
and plasma proximity.
Examples of the successful implementation of the above-
considered procedures and recommendations are presented in
figures 12–16. The plasma rf potential spectra measured in the
11
Topical Review
Figure 13. Two-channel (for the measuring and the reference
probes) rf filter built according to the data shown in figure 12 [55].
Figure 14. Filter impedance versus frequency of the two-channel
filter shown in figure 13 [55].
mid-plane of an argon CCP driven at 13.56 MHz are shown
in figure 12. As seen in figure 12, the symmetrical drive of
CCP ensures a significant reduction in the fundamental (and
supposedly all odd) harmonic of the plasma rf potential that
essentially relaxes the requirements for rf-filter complexity.
A two-channel rf-filter circuit designed for this CCP
and its impedance are shown, respectively, in figures 13
and 14. The measuring probe branch of the filter consists
of sub-miniature (6 mm long, 1.8 mm diameter) chokes self-
resonating at 13.6 and 27.2 MHz. The RP branch of the filter
consists of three resistors (to reduce their total capacitance).
The high resistance of this branch is of no concern to sense
the dc and low-frequency components of the plasma potential
because it is a part of the 10 M voltage divider on the
input of the probe driver shown in figure 15. The filter is
connected to the measuring probe P1 and the ring RP P2 that
is simultaneously acting as a rf shunting electrode connected
to the measuring probe via a small capacitor. The use of a
ring RP provides a sufficiently large surface (for shunting) and
minimal plasma global and local (close to P1 ) disturbance.
The complete probe-measurement diagram is shown in
figure 15. The plasma potential, passing rf filter F and a
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Figure 15. Circuit diagram for EEPF measurement in a

symmetrically driven CCP [55]. The circuit incorporates
compensation for the dc voltage in the electrode rf sheath,
low-frequency noise, rf plasma potential and the external dc
resistance of the rf filter shown in figure 13. The filter (external) dc
resistance is compensated by a current-to-voltage converter (OP1 )
with negative input resistance. Input wave form being amplified by
10 times have opposite polarity than probe voltage.
Figure 16. Second derivative of the probe characteristic
(proportional to EEPF) measured in helium CCP at 13.56 MHz
voltage divider (10 : 1), is fed to the input of the probe driver without the filter, curve 1, and with the filter, curve 2 [55] . Note the
OP2 with a gain of G = 10. The probe span voltage is applied distance between the maximum and zero-crossing points of
to the rf electrodes of CCP, whereas the measuring probe P1 undisturbed Ip , V Te ≈ 5 eV.
is ground via a current-to-voltage converter OP1 having a
negative input resistance equal to the dc resistance of the rf filter
(27 ). The parameters of the probe ramp voltage are shown in
figure 15. To maintain a clean probe surface, the probe is biased
with a high negative voltage (−50 V referenced to its floating
potential) between the probe-voltage ramps. The effect of the
rf-noise suppression with the arrangement described above in
a helium CCP symmetrically driven at 13.56 MHz is shown in
figure 16.
Examples of the EEPF measurements in a CCP are given
in figures 17–20. In figure 17, the measured EEPFs in an
argon CCP driven at 13.56 MHz and 30 mTorr are shown as
functions of the electron total energy (kinetic plus potential) at
different axial positions along the rf electric field [73]. Under
this condition of a strong electron kinetic non-locality [74–76],
the EEPFs measured at different positions of a non-uniform
plasma coincide. The EEPFs measured in the same CCP
running at 300 mTorr and shown in figure 18 demonstrate
mixed behaviour: nonlocal in the elastic energy range of
ε < ε ∗ and local-like in the inelastic energy range of ε > ε ∗ . Figure 17. EEPFs measured at different axial positions in argon
CCP at 30 mTorr [73] . The solid curve corresponds to the CCP
Another example of EEPF measurement demonstrating a CCP
mid-plane, whereas the dotted curve corresponds to a position close
transition in helium gas from the α mode to the γ mode [58, 77] to the plasma boundary.
is shown in figure 19. The change in the electron temperature
and the plasma density corresponding to the measured EEPF
during the mode transition is shown in figure 20. to the skin layer due to the interaction of the rf magnetic
A special situation arises in rf plasmas with a large rf-field field with the rf current [78, 79]. Significant gradients of
gradient, such as in the near rf electrode area of a CCP and the rf electric field arise in rf plasmas immersed in the dc
in the skin layer of an ICP. In the last case, the situation magnetic field, when the rf discharge current induces a rf hole
can be exacerbated in low-frequency ICPs operating in the effect across the plasma [80]. Due to non-uniformity in the
non-linear regime when a significant rf polarization field of discharge rf current (and, sometimes, in the magnetic field),
the second harmonic (much larger than the induction-heating it is difficult to find a sufficiently large rf equipotential area
rf field for the fundamental harmonic) is induced normal to place the shunting probe in the vicinity of the measuring

12
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 18. EEPFs measured at different axial positions in argon
CCP at 300 mTorr [62]. The solid curve corresponds to the CCP
mid-plane, whereas the dotted curve corresponds to a position close
to the plasma boundary. ε∗ denotes the excitation energy.
Figure 19. EEPF evolution during the transition of the helium CCP
into the γ mode [58].
probe. In this case, the biasing of the measuring probe with a
rf voltage equal to the plasma rf potential may be the only
remedy to avoid rf distortions in the EEDF measurement.
This method of mitigating the rf voltage in the probe sheath
requires sophisticated analogue and digital electronics able to
test numerous combinations of phases and amplitudes for each
discharge condition.
13
Topical Review
Figure 20. Plasma and discharge parameter evolution during the
CCP transition to the γ mode [58].
SS OD=1.6 mm
2cm
quartz to dc bias L L
OD=0.5mm 4 2 0.9MHz
r C
to probe L3 4 L 0.45
station 1
z C C C
3 1 2
Figure 21. Probe circuit diagram for EEDF measurement in
strongly non-uniform induction and polarization electric fields in
ICP operating in the non-linear regime [81].
A more simple, convenient and robust way is to bias the
measuring probe with the plasma rf potential sampled with a
dc-biased small RP placed close to the measuring probe. The
circuit diagram of an arrangement used for EEPF measurement
in the skin layer of a low-frequency (0.45 MHz) pancake ICP
operating in the non-linear regime is shown in figure 21 [81].
To increase the coupling between the plasma and the RP
(figure 21, top), the probe is biased to be near the plasma
potential, where the probe sheath has a minimal resistance.
A source of dc current (with high output resistance) must be
used to prevent the shunting of the RP by the dc current source.
The primary winding (L2 ) of the 1 : 1 transformer connected
between the RP and the biasing source induces the rf plasma
potential that rf biases the measuring probe on the secondary
winding (L1 ). The transformer with bifilarly wound, L1
and L2 , is resonated at 0.9 MHz with trimmer capacitor C1
connected to L1 , resulting in an increase in its impedance in
Q ≈ 60 times, thereby acting as a high-impedance rf filter.
Both probes were azimuthally oriented along the induction rf
field, and the EEPF measurements were performed along the
axial direction.
J. Phys. D: Appl. Phys. 44 (2011) 233001
3.4. Time-resolved EEDF
Time-resolved EEDF-measurement techniques have been
considered in reviews [19, 21] and in works cited therein. Since
then, considerable improvement in the probe data acquisition
and digital-processing technique has been achieved. There
are two approaches to the measurement of EEDF in time-
variable plasmas. One is the instant measurement of the
probe characteristic during a probe scan time τs much shorter
than the plasma transient process τp , when the EEDF change
during the measurement is negligible. This method has speed
limitations associated with transient processes in the probe
sheath and surrounding plasma (presheath) and with probe-
sheath capacitance and probe stray capacitance to the plasma
and to the ground.
At low gas pressure defined by the condition (12), the
electron transient time τe across the probe sheath is on the order
of ωe−1 , where ωe is the electron-plasma frequency. Thus, τe
is negligibly small compared with the usual plasma transient
time τp . The ion transient time τi , which is on the order of ωi−1 ,
where ωi is the ion-plasma frequency, is irrelevant to EEDF
measurement because usually d2 Ii /dV 2 d2 Ie /dV 2 . The
real limiting factor affecting speed of the probe measurement
is the displacement currents Ic due to probe-sheath capacitance
(probe tip, and probe holder) to plasma and due to probe lead
capacitance to ground.
The sheath capacitance is non-linear and frequency
dependent and is thus dependent on both V and dV /dt [82, 83].
For a small, time-variable amplitude of the sheath voltage Vt
(V = Vdc + Vt and Vt Vdc ), the displacement current across
the sheath at the floating potential (Vdc = Vfl ≈ (3–5)Te /e)
is equal to the electron current (Ic = Ie ) at ω = (3 − 5)ωi .
At more negative probe voltages and/or higher frequencies,
Ic > Ie and the contribution of d2 Ic /dV 2 may be comparable
to d2 Ie /dV 2 , resulting in EEDF distortion in the inelastic
energy range (ε > eVfl , ε∗ ). A rough criterion for a fast
EEDF measurement unaffected by displacement current can
be written as V −1 dV /dt ωi , corresponding to maximal
EEDF time resolution t = (10–30)ωi−1 . This value for
t is usually much less than the characteristic time of the
plasma parameter transient processes. For example, in the
afterglow stage of a pulsed discharge, the plasma density decay
time, τpN = N −1 dN/dt is τpN ≈ /vB , τpN ≈ 2 /λi vB
and τpN ≈ 2 /Da for ion inertia, variable ion mobility and
diffusion-controlled plasmas, respectively. Here, is the
plasma characteristic size, vB is the ion sound speed, λi is
the ion mean free path and Da is the ambipolar diffusion
coefficient. Note that the values of vB and Da in expressions for
τpN are mainly defined by the cooled electron temperature at
the late stage of afterglow. The electron temperature relaxation
time τpT = Te−1 dTe /dt is always much smaller than τpN and,
for low-pressure discharges, is defined by electron inelastic
collisions (in the early afterglow stage) and by the escape of
electrons to the chamber wall.
Given the non-linear and time-dependent nature of the
sheath capacitance, it seems that there is no chance for an
accurate accounting of the displacement current in the EEDF
measurement. However, the maximal slue rate of the probe
voltage, dV /dt, ensuring no EEDF distortion due to probe
14
Topical Review
1012
Ar, 30 mT, 50 W; off cycle
T = 2 µs
on
T = 20 µs
off
1011
eepf (ev -3/2 cm-3 )
1010
t = 2.8 µs
3.6
4.4
6.8
9.2
12.4
18.8
109
0 5 10 15
electron energy (eV)
Figure 22. EEPF evolution in the afterglow stage of a periodically
pulsed ICP [62]. The position of EEPF from left to right
corresponds afterglow elapse time.
displacement current, can be found experimentally with a
steady-state plasma and variable dV /dt.
A more productive and successful method of time-
resolved EEDF measurement (but only applicable to a time-
periodic plasma, such as a repeatable pulse or low-frequency
discharges) is the collection of many relatively slow probe
scan data with the following rearrangement of the data points
corresponding to the same phase of the transient process. With
the speed and precision (1–10 MHz and 10–16 bit) of the
modern digitizers of digital scopes and processing cards, the
collection of 106 –107 sample points is quite sufficient to obtain
103 –104 data points averaged over 103 probe characteristic
traces. With the probe scan time of 1 ms, 1000 sample
points along each trace and averaging over 1000 traces, the
measurement of EEDF in a large dynamic range (up to 60 dB)
is possible with a resolution on the order of 1 µs.
Time-resolved EEDF measurements according to the
above-described procedure have been recently performed in
different kinds of ICP in pulsed and CW modes [62, 84–86].
The EEPF evolution in the afterglow stage of a periodically
pulsed ICP (T = 22 µs with duty factor 9%) is presented in
figure 22 [62]. Here, seven EEPFs are shown in order from
left to right according to the elapsed time t after the pulse
end, starting at t = 2.8 µs. The shift in the EEPF position
along the energy axis corresponds to the plasma potential shift
referenced to the plasma potential at t = 2.8 µs. The origin
of the EEPF irregularity at t = 2.8 µs (rounding at small
energy and oscillation at higher energy) is not clear. It may
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Ar, CW 100 W 1011 50 kHz, 2 A

p Te
(mT) (eV)

eepf (eV -3 / 2 cm-3)

1
3.0 6.5 1010
electrron temperature (eV)

10 4.2
30 3.1
100 2.1 5 µS
300 1.5
109 t = 1 µS

300
108
0.1 0 1 2 3 4 5 6 7 8
100 energy (eV)
Figure 24. EEPFs measured in inductive hf light source at minimal
30 (t = 1 µs) and maximal (t = 5 µs) instant electron temperature [86].
10
3 mT
80 1.5
1 10 100 1000 50 kHz
time (µs)
lamp voltage (abs. value V)

electron temperature (eV)

Figure 23. Evolution of electron temperature in the afterglow stage
in a periodically pulsed ICP for different argon pressures [62].
be associated with the plasma and/or the ringing effect of the
measuring electronics immediately following the pulse.
The time evolution of the effective electron temperature
Te = 23 ε in the afterglow stage, found by the integration of the
measured EEPFs in a periodically pulsed plasma (T = 2 ms
with duty factor 50%) for different argon gas pressures, is
shown in figure 23 [62]. The steady-state values of the electron
temperature for different argon pressures are also shown. The
electron cooling occurs more rapidly at lower gas pressures.
The main mechanism of the electron cooling is the escape of
electrons to the chamber wall (diffusion cooling) [62, 84, 87].
At 3 mTorr, 60 µs after the pulse, the electron temperature
drops 130-fold from 6.5 eV in the steady state to 0.05 eV, which
is close to room temperature.
EEPFs in a high-frequency (hf) inductive lamp filled with
0.3 Torr of argon and 7 mTorr of mercury and operating at
25–250 kHz were measured with 0.5 µs temporal resolution
at different phases of the hf cycle in [86]. Two EEPFs
corresponding to minimal (t = 1 µs) and to maximal (t =
5 µs) electron temperatures are shown in figure 24 for the
lamp driven at 50 kHz with a 2 A discharge current. In these
conditions, plasma density is essentially time independent
because the hf period is much less than the plasma diffusion
time τpN ≈ 2 /Da . Calculated by the integration of the
measured EEPF, the electron temperature along the hf period
and the lamp voltage along the 60 cm closed discharge path
are shown in figure 25. Note an asymmetry in the Te (t) wave
form, the shift between the electron-temperature extremes and
the corresponding discharge voltage extremes.
2
60
8 1
2A
40
4
0.5
20 8
0 0
0 5 10 15 20
time (µS)
Figure 25. Electron-temperature and lamp-voltage variations during
the rf period [86].
3.5. Ion-current effect
In measurements of EEPF proportional to the second derivative
of the electron current to the probe Ie ≡ d2 Ie /dV 2 , a
differentiation procedure is usually applied to the whole probe
current that consists of the electron and ion components.
Therefore, one may expect an influence of the second derivative
on the ion current Ii ≡ d2 Ii /dV 2 on the measured EEPF. This
question was considered in [88, 89], where the ratio of Ie /Ii
as a function of the normalized probe voltage η = −V /Te was
calculated for radial and orbital collisionless ion motion around
a cylindrical probe for different shapes of EEPF and different
values of a/λD . It has been shown that an appreciable influence
of Ii may occur for thin probes, light gases and large negative
probe voltages when the orbital-motion model is applicable for
the ion current to the probe (a/λD < 1). Under this condition,
when the effect of Ii is maximal, the expression for the ratio

15
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 26. Normalized values of Ie , Ii and Ip (absolute values)
according to expression (16).
Ie /Ii for a Maxwellian EEDF and a/λD 1 was derived as
Ie /Ii = (4πM/m)1/2 η3/2 exp(−η).
The ratio Ie /Ii given by (16) does not depend on a/λD and
can be used as an estimate for the maximal possible Ii effect
in the assumption of a Maxwellian EEDF. Normalized values
of Ie , Ie and Ip = Ie + Ii are shown in figure 26 for different
gases. Assuming the error due to Ii contribution to be within
10%, one can obtain the dynamic range of EEPF measurement
unaffected by Ii in argon plasma as 66 dB (over three orders
of magnitude) from (16) and figure 26. This dynamic range
of EEPF measurements, corresponding to an energy span
of 7.5Te , exceeds the dynamic range of the majority EEPF
measurements in argon plasmas. In EEPF experiments, the
dynamic range of EEPF measurements is limited by the low-
frequency noise (sometimes mistakenly interpreted as a type of
resonance effect in electron kinetics). The effect of Ii on EEPF
measurement is usually considerably smaller than that given
by (16), except for very thin probes, low-density plasmas and
some special EEPF shapes. The effect of Ii on EEPF manifests
itself as a slowdown in the drop of the function ln[Ip (V )], at
large negative probe voltages seen in figures 26 and 27.
Both electron and ion currents to the probe (and their
derivatives) essentially depend on the shape of EEPF [90].
The authors of [89] calculated and compared the values of
Ie and Ii for non-Maxwellian EEPFs for orbital and radial
ion motion. They considered a class of EEPFs proportional
to Cq exp[−(ε/εq )q ], where q = 0.5, 1, 2, 3; Cq (q) and εq (q)
are coefficients depending on q. The calculations for radial ion
motion were performed for a cylindrical probe at a/λD = 1, 10
Topical Review
(16)
Figure 27. Normalized values of Ie , Ii and Ip (absolute values) for
the radial motion of protons at different ratios of a/λD = 0.01
(upper curves), 0.1 (middle curves) and 1 (lower curves) for
Maxwellian (solid curves) and Druyvesteyn (dashed curves)
distributions. For other ions, Ii is (M/MH )1/2 times less [89].
integral of the EEDF and usually referred to as the electron
temperature for non-Maxwellian EEDF.
As seen in figure 27, at large electron energy, Ii for
Druyvesteyn distribution is about twice as large as that for the
Maxwellian distribution, whereas the opposite is true for the
Ie values. Therefore, the dynamic range and energy span of
EEPF (where Ii can be neglected) for Druyvesteyn distribution
are less than that for Maxwellian distribution. For EEPF with
q = 0.5, or for a bi-Maxwellian distribution typical in low-
pressure CCP, the Ii effect is less than that for a Maxwellian
distribution [89]. Shown in figure 27, the values of Ii for a
more realistic case of a/λD  1 are even less than that shown
in figure 26 for the limiting case of a/λD 1. The Ii values
for non-hydrogen gases are (M/MH )1/2 times smaller than that
given in figure 27. Therefore, in the majority of cases of EEPF
measurement with a noise-limited dynamic range 50–70 dB,
the effect of Ii can be neglected.
A popular method (implemented in many commercial
probe-diagnostic instruments) to eliminate Ii is the subtraction
from the total probe current Ip (V ) of the ion current Ii (V )
extrapolated from a large negative probe voltage (where
and 100. The normalized values of Ie , Ii and Ip as functions of
the parameter η = −V /Teff are given in figure 27 for hydrogen
ions at Maxwellian (q = 1) and Druyvesteyn (q = 2) electron-
energy distributions and different a/λD . Here, Teff = 2/3ε
is the effective electron temperature found as a corresponding
16
Ii  Ie ) to a lower probe voltage (where Ii and Ie are
comparable). This procedure for the separation of the
electron current (quite reasonable and justified in the classical
Langmuir-probe routine) is (in our opinion) meaningless for
the EEPF measurement and may even introduce more error.
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 28. The procedure to correct the EEPF measurements
affected by the ion current [12, 94].
There are at least two reasons for this negative opinion. First,
the extrapolation procedure is generally ambiguous. Second,
the existing ion current Ii (V ) theories are not accurate enough
to justify their application to account for the Ii effect on
EEPF. Suffice to say, the basic plasma parameters calculated
according to these theories may be in error up to an order
of magnitude [91–93]. The reason for the inaccuracy of the
ion-current theories is that they do not account for many non-
ideal conditions of the real plasma experiment, such as non-
Maxwellian EEDF, unknown ratio of Ti /Te , rare ion collisions
that destroy ion orbital motion, ion drift velocity approaching
Topical Review
Figure 29. Influence of probe-surface contamination on Ip
(absolute values) in glow discharge in a mixture of neon and a 5%
addition of benzyl. pR = 1.2 Torr cm, 20 mA discharge
current [12, 97] . Hot probe (1), probe after heating termination at
time 0 (2), 30 s (3), 60 s (4) and 120 s (5).
(in figure 4, P1 and P2 ). Note that the two-probe technique
has limitations at high gas pressures, when the inequality
a[ln(π l/4a)] λe is not satisfied for the large probe. More
information on the use of this technique in different gas-
discharge plasmas can be found in [12].
3.6. Probe-surface effects
Probe-surface contamination is a long-recognized problem in
probe diagnostics that has been studied in many works and is
summarized in [10, 12, 19, 21, 25]. Therefore, we will briefly
vB  (Ti /M)1/2 caused by ambipolar field and failure of one-
dimensional cylindrical sheath structures at highly negative
probe voltages. At highly negative voltages, Ip (V ) can almost
always be well fitted to a function proportional to (−V )1/2 .
However, in contrast to popular belief, this fit does not prove the
validity of the orbital-motion theory for the extrapolated Ii (V )
in a particular probe experiment. After all, due to the error-
augmentation effect inherent to the differentiation procedure,
a small error in Ip (V ) can lead to enormous error in Ii .
A practical way to minimize Ii error in EEPF
measurement is the subtraction from the measured Ip of the
value of Ip extrapolated from large probe voltage [94], where
Ip is apparently dominated by Ii , as shown in figure 28. Of
course, this technique has limited applicability and works in
the electron-energy interval where Ie and Ii are comparable.
Measurements with two probes having different radii allow
the avoidance of the influence of Ii on the measurement of
EEPF [95, 96]. With this technique, the small probe provides
accurate data for the low-energy part of the EEPF, whereas
describe probe contamination and its effect on the quality of the
measured EEPF, stressing the possible solutions to the problem
with examples of EEPF measurements in chemically active
plasmas.
Probe-surface contamination occurs in laboratory noble-
gas plasma due to residual gases and, to a much larger extent,
in plasma-chemical reactors with an intensive deposition of
reaction products to the probe. The presence of a low-
conductivity layer on the probe surface leads to a change in
the probe work function and to an additional surface resistivity,
contributing to the probe circuit resistance. The contaminated
probe surface may also result in an increase in secondary
electron–electron, electron–ion and electron photoemission.
All of these effects may bring significant distortions into the
measured EEPF and especially into its low-energy part.
An example of a dramatic probe-contamination effect on
Ip is shown in figure 29 for a dc glow discharge in argon
with 5% of benzyl at a different time after probe heating at
the thick probe (with a/λD  1) has a negligible Ii at large
negative probe voltages, corresponding to the high-energy part
of the EEPF. In the latter case, due to large Ip and Ip signals, a
relative reduction in internal apparatus noise can be achieved.
A problem in the described method may arise from the
inability to obtain accurate measurements with the thick probe
of the plasma potential needed to adjust the twice-measured
EEPF. In this case, the adjustment of the two EEPFs can be
performed in the middle-energy range where both probes give
accurate measurement of Ie .
A way to implement the two-probe technique could be
to interchange the roles of the measuring and large RPs
17
900 ◦ C [97]. The change in time of the voltage drop across
the probe-contamination layer in pure argon-glow discharge is
shown in figure 30 [98].
The change in the probe work function is a real problem
in many published EEPF measurements, even in discharges in
noble gases. Estimation shows that even at very low levels of
residual gases, corresponding to a gas pressure of 10−6 Torr,
a monolayer of residual gases appears on the probe surface in
less than 1 min. It has been found that the probe work function
is very sensitive to the probe temperature. The change in the
probe current causes a temperature drift and, consequently,
a change in its work function. This change results in probe
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 30. Variation of the effective probe work function in argon
plasma. The probe material is molybdenum (1), tantalum (2) or
tungsten (3) [98].
characteristic distortion because the probe-sheath voltage does
not correspond to the applied probe voltage. As shown in [99],
the change in the probe work function of the probe in mercury
plasma can reach 1.5 V (close to the electron temperature in this
experiment), with a minimal time constant τT of approximately
2 s. The change in the work function during the probe scan
can be seen in the hysteresis of the probe characteristics and
its derivatives obtained for different directions of the probe
voltage change when the scan time τs is comparable to the
probe temperature time constant τT . The hysteresis disappears
Topical Review
or with sputtered metal from other electrodes (such as rf
electrodes in a low-pressure CCP) and from the probe itself
during its cleaning by ion bombardment. A relatively simple
solution to this problem can be a thin dielectric sleeve
preventing electrical contact between the probe and its holder,
as shown in figure 2.
Incorporating the considered above remedies to address
the problems associated with the probe circuit resistance,
low-frequency and rf plasma noise and change in probe
work functions together with advanced analogue and digital
electronic acquisition and processing routines in the probe-
diagnostics systems described in [51, 55] resulted in the high-
quality EEPF measurements shown in figures 8, 9, 11, 17,
18 and 19. However, the probe systems developed for basic
plasma research and their commercial imitations still are not
wholly suitable for reliable EEPF measurements in commercial
and laboratory plasma-chemical reactors. Commercial plasma
reactors (and some laboratory reactors) are not designed for
plasma diagnostics, and attempts to directly apply probe
instruments, proved in basic plasma experiments, are often
frustrating.
There are many problems with implementing meaningful
probe diagnostics in commercial plasma reactors: (a)
large plasma rf potentials corresponding to the plasma
source and rf-bias fundamental frequencies and their
harmonics; (b) plasma low-frequency instability; (c) high
rate of reaction-product deposition contaminating the probe
in very slow (τs  τT ) and at very fast probe scans. In the
first case, the probe temperature (thus, its work function) is in
equilibrium with the probe voltage and the probe characteristic
and its derivatives are distorted. In the second case, due
to thermal inertia, the probe work function remains constant
during the probe scan and thus does not distort the probe
characteristic and its derivatives.
It has been known for decades that the radical remedy to
avoid distortion caused by changes in the probe work function
is a fast (τs τT ) probe-voltage scan that can be achieved
using a pulsed-probe technique with a scan-time order of
ms [19]. To clean the probe between pulses of the scan
voltage, a large negative (or positive) bias voltage is applied
to the probe, as shown in figure 15. For a negatively biased
probe, the cleaning is due to ion bombardment, whereas for a
positively biased probe, the cleaning is due to probe heating by
the electron current. Note that a fast-pulsed probe technique
for EEPF measurement was used in the 1970s but has been
mostly abandoned.
The cleaning of the probe by biasing should be done
cautiously. The highly negative probe bias leads to probe
sputtering, changing it size, especially for thin probes and
dense plasmas. Therefore, in the pulsed repetitive regime,
the negative probe bias (referenced to the plasma potential)
should not exceed 50–100 V. Probe heating by electron current
has the advantage of preventing polymer-film deposition on the
probe, but the probe may accidentally be melted with excessive
probe current. For this reason, hot probes are rarely used in
experiments.
In some cases, the probe and its dielectric holder may
be contaminated with a conductive layer of reaction products
18
surface, sometimes preventing even classical Langmuir-probe
diagnostics; and (d) overly high impedance between the plasma
and grounded chamber (mostly due to the limited surface of the
conductive chamber area, its contamination, and/or dielectric
protective coating).
Today, commercial plasma-simulation codes are the main
tool for the study of plasma electrodynamics, transport and
kinetics in commercial plasma reactors. These codes, when
applied to plasmas in complex processing gas mixtures that
involve molecular and electronegative gases and particles,
are missing many cross sections for some plasma-chemical
reactions. They also are missing some important effects of
nonlocal and non-linear plasma electrodynamics that may even
be dominant in rf plasma at low gas pressure [100]. In this
situation, reliable EEDF measurements would give valuable
experimental data for understanding the intricate processes in
commercial rf plasma reactors and for validity testing of the
corresponding numerical codes.
In an attempt to address the specific problems of EEPF
measurements in chemically active plasma reactors with large
amplitudes, wide spectra of rf noise and high deposition rate,
the VGPS probe-diagnostic system [60] has been developed
and tested in laboratory and commercial plasma reactors.
Incorporating an automatic fast probe cleaning involving
ion bombardment, rf biasing and forced probe heating by
electron current allows for the reliable measurement of EEPF
in plasma reactors with high deposition rates and significant
noise.
An example of this measurement [101] given in figure 31
shows a snapshot of EEPF and plasma parameters from
the display of the VGPS instrument. The measurement
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 31. EEPF and plasma parameters measured in the
microcrystalline-silicon-deposition plasma in a 2.45 GHz
matrix-distributed ECR reactor [101].
was performed in a typical microcrystalline-silicon-deposition
plasma in a 2.45 GHz matrix-distributed ECR reactor [102]
driven at 1.5 kW and 10 mTorr of a gas mixture consisting of
Ar, SiF4 and H2 with corresponding flow rates of 100, 15 and
150 sccm. Under these conditions, the deposition rate in the
probe vicinity is 0.3 nm s−1 and the conventional Langmuir-
probe measurement (without fast probe scan and continuous
probe cleaning) has been considered to be impossible.
The EEPF displayed in figure 31 has three distinguishable
parts. The bulk of the distribution, in the electron-energy
interval between 0 and 2 eV and EEPF interval of 60 dB,
corresponds to a Maxwellian distribution with an electron
temperature Te = 0.33 eV. Starting with ε = 2 eV, the EEPF
slope changes (most likely due to the contribution of the
second derivative of the ion current, mainly consistent of light
hydrogen ions). After 3 eV, the EEPF sharply decreases due
to the limited dynamic range of the measurement instrument.
By extrapolating Ip (V ) at large energy (assuming Ip = Ii )
to lower energy and subtracting it from the measured Ip (V ),
one can recover the EEPF in the energy range between 2 and
3 eV, thus extending the energy interval by 0.5–1 eV and the
dynamic range of the recovered EEPF by 6–10 dB. The part of
the EEPF with ε > 3 eV must be considered noise and removed
from the EEPF.
Another example of the measurement of EEPF in an ICP
plasma reactor with low-frequency instability and deposition
of polymer layers in a 30 mTorr gas mixture of H2 /CF4 is shown
in figure 32 [103]. As one can see in figure 32, a measurement
of EEPF with a dynamic range about 50 dB corresponding to
the electron-energy interval of 7Te is quite possible, despite
the relatively high level of deposition and plasma instability.
19
Topical Review
Figure 32. EEPF measured in the H2 /CH4 ICP reactor with
polymer-layer deposition at different contents of CH4 [103].
3.7. Instrumental functions in probe measurements
In practice, because of various systematic errors, the measured
EEPF fpm (ε) is always somewhat different from the real,
nondistorted EEPF fp (ε) in plasma. Previously, we discussed
how to design probe experiments to reduce the difference
between fpm (ε) and fp (ε). However, systematic errors
cannot be completely excluded. The influence of some
spurious instrumental effects on the probe measurements of
the EEPF can be described by applying so-called instrumental
(or apparatus) functions [14, 21, 23, 25]. The result of
measurements of the EEPF fpm (ε) is a convolution of the real
EEPF fp (ε) and the instrumental function A:

∞
fpm = fp (x)A(ε − x) dx ≡ fp ∗ A. (17)
−∞
If there are n distortion effects, described by corresponding
instrumental functions A1 , A2 , ..., An , then the function A is a
convolution A1 ∗A2 ∗...∗An and gives the full description of all
included effects. As we see in equation (17), the instrumental
function is reproduced by the measurement system when fp (ε)
is a narrow (ideally δ-function) peak at a certain energy. The
instrumental function shows how this peak is widened and
distorted during the measurements. Accordingly, the distortion
of the EEPFs of arbitrary form can be accounted for with
equation (17).
Here are some important application features of the
instrumental function that must be considered in processing
probe measurements:
(1) The instrumental function allows the description of a
number of (but not all) distortion effects by uniformly
and conveniently taking them into account.
(2) The instrumental function A gives a full description of
the distortion of the real EEPF by the included distortion
effects. The correct result of measurements can be found
as a solution of equation (17) (however, this problem
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

is ill-posed and may thus be accompanied by error

augmentation, similarly to differentiation procedures).
(3) If the width of the instrumental function is finite and
essentially smaller than an interval of substantial change
in the EEPF fp (ε), the latter is well reproduced by fpm (ε).
(4) For a Maxwellian EEPF, fpm (ε) coincides with fp (ε),
with the exception of the energies from 0 to the energies
corresponding to the width of the instrumental function.
(5) Instrumental functions allow the comparison of measure-
ment methods and the corresponding equipment and, thus,
the choice of a method providing minimal distortion.
(6) If the EEPF has the property

ε1
ε2 −ε
√ √
fp (ε) ε dε + fp (ε) ε dε
ε1 −ε ε2 Figure 33. A circuit for simulating a narrow-energy group of

ε2 electrons in a plasma. D is a diode, B is a dc source, R is a resistor,
√
fp (ε) ε dε, (18) P is a probe and M is the measurement system. Switches S1 and S2
ε1 allow the probe circuit and/or the diode circuit to be connected to
where ε is the half-width of the instrumental function and the measurement system M [111].
ε1 − ε  0, then the electron density in the energy interval
from ε1 to ε2 can be found directly from fpm (ε) even if fp (ε) is
significantly distorted (e.g. when fp (ε) has a sharp maximum).
In this case, the form of the maximum can be strongly distorted
during the measurements, but the area under the maximum still
gives the correct electron density.
The various instrumental functions and their calculations
for the Druyvesteyn method have been described in
the literature [14, 21, 25, 104, 105]. Some describe the
measurement methods of d2 Ip /dV 2 [48, 49, 106], whereas
others illustrate the influence of the finite measurement time
[21, 106]. A third group accounts for reflection and secondary
emission of electrons from the probe surface [107, 108], and
a fourth group accounts for oscillations of plasma potential
[109]. As an example of practical application of instrumental Figure 34. The measured instrumental function of the
functions, in [110] different methods of differentiation to SMARTProbe®/ (1). The same function in the presence of potential
obtain d2 Ip /dV 2 are compared. oscillations with an amplitude of 2.5 V (2) [111].
The instrumental function of a d2 Ip /dV 2 measurement
arrangement can be found by probe substitution with an (decrease the energy resolution) but do not change the system
electrical circuit having the second derivative of its volt/ampere sensitivity (the area under measured curves remains the same).
characteristic close to a δ-function. An example of such The above results are in good agreement with theoretical pre-
a circuit used in the measurement of instrumental function dictions [111].
in the commercial probe system SMARTProbe® is shown To demonstrate how to determine the measurement
in figure 33. The measured instrumental function A of sensitivity with real plasma, the probe measurements of
SMARTProbe®, working in the boxcar mode with a voltage d2 Ip /dV 2 obtained in the ICP afterglow plasma are shown
sweep of 40 V, 1000 samples per point, 10 sweeps per scan and in figure 35. The diode circuit has been included in the
the voltage step of differentiation of 4 V, is shown in figure 34. measurements, as shown in figure 33, via switch S2 . In this
In this example, the area under the measured d2 Ip /dV 2 is equal case, the sweep voltage was set to 10 V with 100 samples
to 4.5×10−6 A V−1 , which defines the sensitivity of the system. per point and 10 sweeps. For those conditions, the electron
The energy resolution of the system is about 4 eV. Note that temperature in the plasma is 0.2 eV. With S1 ‘open’ to remove
the instrumental function found in this way characterizes the the diode circuit, no additional peaks are observed in the
instrument but not the probe itself, which can be affected by measured d2 Ip /dV 2 (dots). After connecting the diode circuit
an internal probe circuit resistance, probe contamination and with a dc source B of 7.3 V, an additional maximum is seen
secondary emission or electron reflection on the probe surface. in d2 Ip /dV 2 (solid line). Changing the R value allows for the
The combined instrumental function due to variation of the area under this maximum. In figure 35, the
SMARTProbe® system convolution effect and that caused additional artificial maximum on the d2 Ip /dV 2 curve is shown
by the artificially introduced probe rf potential is shown in by an arrow. In this case, the artificially created maximum
figure 34. The rf voltage with an amplitude of 2.5 V was in d2 Ip /dV 2 (with R = 220 k) gives an artificial ‘density’
introduced into the circuit shown in figure 33. As expected, the Ne = 5×106 cm−3 (note there is a misprint in [111] for Ne and
rf oscillations increase the width of the instrumental function the system sensitivity). Increasing R allows the determination

20
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Figure 36. Instrumental function A(ε) measured in a

neon-afterglow plasma (1). The calculated function for the ‘clean’
probe (2). The calculated function for a probe with electron
reflection at reflection coefficients of 1 − 0.016 V−1 (3) and
1 − 0.056 V−1 (4) [108].

Figure 35. Measured EEDs in argon-rf-afterglow plasma without

(dots) and with (solid line) an additional artificial maximum
(indicated by arrow). The curves coincide for slow electrons. The
gas pressure is 30 mTorr, the repetition frequency is 400 Hz, and the
time after current interruption is 0.7 ms [111].

of the ultimate sensitivity of the measurement device. In

this particular case, it is 2 × 106 cm−3 . Thus, the simple
circuit consisting of a diode, a resistor and a dc voltage source
allows the determination of the instrumental function and the
sensitivity of the particular differentiation instrument for a
more detailed characterization of a probe-measurement device.
Another method of measuring instrumental functions
includes all effects associated with the measuring system, the
effects associated with probe itself and different distortion
effects considered previously. For this purpose, the EEPF in
a plasma with mono-energetic (but isotropically distributed)
electrons should be measured [108]. One example of such
a plasma is an afterglow plasma. In this case, a group of
mono-energetic electrons arises in plasma-chemical reactions,
such as Penning ionization of excited metastable atoms by
collisions with other (excited) atoms. The results of such
measurements in neon and xenon afterglows are shown in
figures 36 and 37. In figure 36, the measurements were
performed with a clean molybdenum probe. The cleaning
was performed by ion bombardment of the probe surface.
In figure 36, the measured instrumental function (solid line)
is very close to the theoretical one (dashed line 2), which
indicates that the reflection and/or secondary emission of
the electrons from a clean probe surface are negligible. In
the same figure, the theoretical curves with variation of the
reflection coefficient are also shown. Figure 37 shows similar
measurements with the probe having not been cleaned for a
significant amount of time. The reflection and/or secondary
emission are essential for dirty probes, producing negative
parts of the instrumental function. These measurements
demonstrate a harmful effect of reflection and/or secondary
emission on the EEPF measurements, again confirming the
necessity of probe cleaning during EEPF measurements even
in noble-gas plasmas.
Figure 37. An instrumental function obtained from a probe with a
dirty surface [108].
4. More complex plasma: beyond the limitations of
the Druyvesteyn method
In basic plasma research and in some industrial applications,
there are situations where the use of the Druyvesteyn formula
in its classical form is not valid. The following section
describes EDF-measurement methods in plasmas with higher
gas pressures, substantial magnetic fields and anisotropy. The
elements of electron-spectroscopy analysis will be discussed
at the end of the section.
The currently existing probe theories allow, in principle,
EDF measurements for the above conditions. However,
measurement techniques based on these theories have not
found broad applications. As a result, these techniques have
not been studied to the same extent as the Druyvesteyn method;
they are less developed, and their complete verification has not
yet been completed. Therefore, this review cannot provide the
same completeness in technical details for the measurements.
Nevertheless, we will present the results of available efforts
on EDF measurement in non-traditional (for the Druyvesteyn

21
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

method) plasmas to demonstrate their potential value and

to challenge our colleagues to exert more effort into the
development of EDF-measurement techniques applicable to
these plasmas.

4.1. Higher pressures (plasmas with near-probe collisions)

In the case of elevated gas pressures, when electron collisions
with other plasma particles in the near-probe plasma region
become essential and inequality (12) is not satisfied, the
Druyvesteyn formula is, strictly speaking, invalid. The case
when λe ≈ a has been considered in [10, 114, 115]. It has been
shown that distortions inflicted by collisions in the measured
EEPF are more important for its low-energy part and that they
can be accounted for by the application of a regularization
procedure. The regularization formalism has been developed
in [114] for a spherical probe. For a more practical cylindrical
probe, the following integral equation, coupling the measured
EEPF fpm (ε) with the real EEPF undistorted by collisions, Figure 38. EEPF in a helium afterglow at different times τ with a
fp (ε), has been derived in [115] probe having a = 0.1 mm, l = 15 mm and a gas pressure of 40 Torr:

∞  τ = 100 µs, Te = 0.047 eV (1); τ = 300 µs, Te = 0.044 eV (2); and
s fp (x)dx τ = 500 µs, Te = 0.043 eV (3) [120].
fpm (ε) = fp (ε) 1 − 2 ,
ε xfp (ε)[1 + s (1 − ε/x)]3
(19) This method simply requires the measurement of dIp /dV
where s is the sink parameter and therefore may be referred to as the first-derivative method.
If sheath is thick but vDe is not a constant, then the simple
4a l formula (23) generally cannot be used and it is necessary
s ≈ ln . (20)
3λe 2a to know the potential profile in the probe sheath, which
Finding the true EEPF, fp (ε) requires a deconvolution can be found in principle from the Poisson equation jointly
procedure with equations (19) and (20) using an iteration solved with the equation of motion of the ions in the sheath.
technique. For a weak collision effect, when s 1, the Formula (23) has been used in practical measurements of
true values of plasma density N and electron temperature Te the EEDFs, for example, in [116–120]. An example of the
can be found through integration of the measured fpm (ε) (to get measurements is shown in figure 38 [120].
Nm and Tem ) and correction according to expressions A more general nonlocal probe theory, which is valid for
N ≈ Nm (1 + 4s /3) (21)
and
Te ≈ Tem (1 − s /2). (22)
For higher pressures, when λe a[ln(π l/4a)] λε , the
problem was solved in [116]. Here, λε is an electron-energy
relaxation length (more details on the energy relaxation length
can be found, e.g., in [25]). The theory [116] is valid for
gas pressures when there are many collisions in the plasma
volume disturbed by the probe (λe a[ln(π l/4a)]), but
electrons from undisturbed plasma reach the probe surface
practically without the loss of the total (kinetic and potential)
energy due to collisions (a[ln(π l/4a)] λε ). For atomic
gases and weakly ionized plasma, λε is on the order of 100λe ,
and the theory thus allows for the expansion of the pressure
range for probe measurements of the EEPF for typical probe
dimensions up to several hundred Torr. It was demonstrated
[116] that for thin (with respect to the probe radius) probe
sheaths (sufficiently high electron density) or arbitrarily thick
sheaths and vDe = const or λe = const (e.g. in argon plasma),
the following formula can be used to infer the EEPF:
3m2 a ln(π l/4a) dIe
f0 (ε = eV ) = − . (23)
8πλe (ε=eV )V e3 Sp dV
Here, De is the electron diffusion coefficient.
22
λε  a[ln(π l/4a)] and thus includes both collisionless and
collision cases, was developed in [120, 121]. The developed
theory gives expressions that are somewhat similar to (19) and
(20). In this case, the electron current to the probe is


8π eSp ∞ (ε − eV )f0 (ε) dε
Ie = . (24)
3m2 eV γ + (1 − eV /ε)(ε, V )
In this expression,

√
1 π l/4 εDe (ε) dr
(ε, V ) = √ ,
λe a (r/a) ε − eϕ(r)De (ε − eϕ(r))
(25)
ϕ(r) is the plasma potential at distance r from the probe and
γ is a factor of an order of unity which can be approximately
given as γ = 4/3−0.62 exp[−λe /(2a)] [25]. In equation (24)
f0 is the EDF in the undisturbed plasma.
For a thin probe sheath or arbitrarily thick sheath and
vDe = const,  = λae ln 4a πl
. Apparently, this method is more
difficult to implement because it requires the solution of the
integral equation (24) to infer the EEPF from the measured
electron probe current Ie . In practice, however, it may be
more convenient to use the Druyvesteyn or the first-derivative
methods because it is always possible to adjust the probe radius
to satisfy the corresponding validity condition.
More details of the general nonlocal kinetic theory can
be found in [21, 22, 24, 25, 27, 120]. In [21], an evaluation
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 39. Calculated ln(−Ie (right) and ln(−Ie (ε)/ε) (left) for a Maxwellian EEPF,  = 1 (1),  = 5 (2),  = 20 (3),  = 0.3 (4),
 = 1 (5),  = 2 (6), and the model Maxwellian EEPF (dashed line) [21].
of the validity limits of the Druyvesteyn and first-derivative
methods based on the general theory has been performed.
Many examples of EEDF distortion with the evaluation of
measurement errors have been analysed. These errors depend
on the specific form of the EEDFs. As an example, the results
of calculations of electron probe current for Maxwellian EEPF,
demonstrating deviation of the solution from ideal (a straight
line), are shown in figure 39. It was shown that, for Maxwellian
EEPF and  > 5, an error in the obtained values of N and
Te using (23) is not more than 10%. Generally, it is possible
to conclude [21] that the Druyvesteyn method is reasonably
Topical Review
4.2. Magnetic fields
The measured EEPF in plasma with magnetic fields is of a
great interest due to the importance of the magnetized plasma
in laboratory and practical applications. In this case, for
theoretical description of electron probe current, the electron
Larmor radius RLe is an important plasma parameter. The
probe current also depends on the probe orientation with
respect to the magnetic lines. The most interesting are two-
probe orientations: magnetic field B is parallel to the probe
and B is perpendicular to the probe. In the presence of weak
applicable if λe > 0.75a ln[πl/4a] and formula (23) may be
used if λe < (a/7) ln[πl/4a].
For practical realization of the first derivative and more
general nonlocal methods, the precautions to avoid the probe
measurements errors considered in section 3 are relevant. It is
clear that probe circuit resistance, probe-sheath rf oscillation
and surface effects can distort the EDF measurement unless
they are properly addressed in the probe measurement. For
magnetic fields (RLe  l, for parallel probe and RLe  a, for
perpendicular probe) the Druyvesteyn method is still valid and
can be used without correction [5]. Increasing the magnetic
field is somewhat similar to increasing the gas pressure;
therefore, an approach similar to that described in the previous
section can be used [21, 22, 24, 25, 27, 116].
It has been demonstrated that in a strong magnetic field
(RLe l, B  l) and (RLe a, B ⊥ l) for thin probe sheath
higher pressures (and magnetic fields described in the next
subsection), it is unlikely to satisfy inequality (12) for the
probe holder and it may create significant disturbances of the
surrounding probe plasma. Therefore, the requirements for a
good probe design may be even more important. Numerical
analysis (similar to that done in [56]) and/or experiments
with different sized probes [112] could contribute to the
solution of the problem but require significant additional work.
Comparative measurements with different probe and probe-
holder sizes and orientations could be useful and are highly
desirable to understand the proper probe design in collisional
and magnetized plasmas.
An additional issue is the temporal resolution of the probe
measurement at relatively high pressures and in the presence
of a magnetic field. In those cases, the transport of electrons
to the probe is slow due to the reduction by the electron
diffusion caused by collisions and the magnetic field. This
issue poses an additional restriction on the speed of the probe
measurement [25]. The timescale for electron diffusion to the
probe is τD = 2p /De , which can be much larger than that
defined by the probe-sheath capacitance for the collisionless
probe regime. Here, p is the diffusion length of the plasma
area distorted by the probe.
23
(a/λD  1) the following formulae are applicable:
for parallel probe (B  l) [21, 25, 116, 122]
3ωH m5/2 dIe
f0 (ε) = − √ , (26)
64 2π e2 a(eV )3/2 dV
for perpendicular probe (B ⊥ l) [25, 122]
3m2 ln(π l/4a) dIe
f0 (ε) = − . (27)
16π 2 e3 V RLe dV
Here, ωH is the electron cyclotron frequency. Figure 40
shows the first derivative to the probe characteristic, dIp /dV ,
measured with a parallel and a perpendicular probe under
the same plasma conditions [122]. Measurements have been
performed in the toroidal plasma device ‘Blaaman’ [123].
It is possible to see that dIp /dV measured by the parallel
probe is essentially smaller than that for a perpendicular probe.
Figure 41 shows the EEDFs measured in [124].
The general nonlocal kinetic theory, similar to that
discussed for higher pressure plasma in the previous section,
has also been developed for the magnetized plasma diagnostics
in [120, 121]. It gives exactly the same expression (24) for
J. Phys. D: Appl. Phys. 44 (2011) 233001
Figure 40. Measured (dots) and calculated (line) Ie in a magnetized
plasma for perpendicular (1) and parallel (2) probes [122].
Figure 41. The EEPF obtained by a perpendicular (a) and parallel
(b) probes (solid curves) and the CASTOR tokamak edge plasma.
Magnetic field B = 1.3 T. The dashed—dotted lines are the
bi-Maxwellian approximation. The dashed line represents the
distribution of the low-energy electron population; the dotted line is
that of the high-energy population. The dashed—dotted line is the
sum of the dotted and dashed lines [124].
electron probe current with ⊥ = a ln(π l/4a)/γ RLe for the
perpendicular probe and  = πl/4γ RLe for the parallel
probe. This theory has been also applied to the interpretation
of the measurements (e.g. in [124, 125]).
24
Topical Review
However, the accuracy of this theory is still unclear and
requires additional studies. For example, the theory uses
the same approach to a boundary conditions as that for the
above higher pressure probes. However, the trajectories of
electrons between collisions are not straight lines but arcs,
which could be a source of unknown error. Therefore, it is
probably more reasonable to use the Druyvesteyn or first-
derivative methods in practice because it is almost always
possible to adjust the probe radius and use different probe
orientations to satisfy the corresponding validity condition. An
analysis similar to that performed for the high-pressure probe
suggests that the Druyvesteyn method is reasonably applicable
if ⊥ ,  < 0.75 and formulae (26) and (27) if ⊥ ,  > 7.
For practical measurements, the above remarks from
the previous subsection should be considered. Because
the diffusion of charged particles to the probe holder can
be essential, it is advisable to place it perpendicular to the
magnetic field. Additionally, for magnetized plasma, the
3a 2 ω2
timescale of diffusion to the probe is of the order of 8v2 νH . This
expression gives, for example, the following restriction on the
application of a fast-sweep probe [113] for the exploration of
magnetized plasma oscillations [25]:
3a 2 ωH2
| eVsw | τsw τfl , (28)
8v 2 νTe
where Vsw is the voltage range of the sweep, τsw is the sweep
time and τfl is the typical fluctuation time. In real experiments,
taking into account that the probe should be fast enough to
resolve oscillations, this condition may be difficult to satisfy.
4.3. Anisotropy
The issue of the EDF in anisotropic plasmas is reviewed in
[19, 21, 25, 27]. In this subsection, we give an abbreviated
review of the methods for evaluation of an anisotropic EDF,
which can be useful for probe diagnostics with very low gas
pressures and/or strong electric fields. Unfortunately, a full
recovery of the anisotropic EDF requires a rather complicated
experimental probe design (flat one-sided probe placed in an
anisotropic plasma under a number of different angles known
with high precision) and a cumbersome numerical recovery
procedure. Therefore, the wide application of a flat one-
sided probe is questionable. Sometimes, it is possible to
use a simple cylindrical probe placed in the plasma under a
number of angles, which can give an insight into the degree
of EDF anisotropy and provide some details of the EDF. A
spherical probe can still measure the EEDF (but not the full
EDF) in the anisotropic plasma, providing that some additional
conditions are satisfied. Of course, information about the angle
distribution in such a measurement is completely lost.
It has been demonstrated [5, 25] that, for the spherical
probe, the Druyvesteyn method in collisionless anisotropic
plasma is valid in two cases: the probe sheath is thin (which
implies sufficiently high electron density) and the probe sheath
is arbitrarily thick but spherically symmetric. Note here that
the spherically symmetric sheath occurs in partially anisotropic
plasma when the plasma contains a high-energy low-density
flux of electrons with an isotropic electron background. The
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Figure 42. The EEPF in a low-pressure (0.1 Torr) hydrogen

constricted arc plasma at the discharge axis.

negligible effect of the fast electrons on the probe-sheath

geometry is due to the fact that the Debay length and Figure 43. Ip with respect to the potential V measured at the
potential distribution in the probe sheath are defined by the discharge axis at a distance Z from the cathode by probes in two
low-energy part of the EDF [90]. The measurements of mutually perpendicular orientations. At Z > 2 mm, I  is the same
for both probes and is proportional to EEPF. The helium pressure is
an anisotropic EEPF with a spherical-probe method were 2.3 Torr, the discharge current is 0.5 A, the electron density is
performed in the study of constricted discharges [126, 127]. 1.1 × 1011 cm−3 [30].
Figure 42 represents the EEPF obtained in a low-pressure
(13 Pa) hydrogen constricted arc discharge. The two peaks (dotted curve) and parallel (solid curve) to the discharge axis
in the measured EEPF at 22 and 35 eV are due to electron at a distance z from the cathode are shown in figure 43 [30].
acceleration by a double-layer structure in the constricted For z > 2 mm, the curves practically coincide, and Ip is
region of the plasma [126].
proportional to fp . As one can see in figure 43, the relaxation
Measurements with cylindrical probes provide some
of the cathode beam (no difference in the measurements made
information about anisotropic EDF [128, 129]. In this case,
by the differently oriented probes) occurs at the distance of
rotation of the probe perpendicular to its axis can actually
z > 2 mm. The method considered above can be used for
indicate the presence of the anisotropy. Due to the symmetry
EDF diagnostics in axially symmetrical partially anisotropic
of the spherical functions Yjk , measurement with a cylindrical
plasmas where the probe sheath satisfies cylindrical symmetry.
probe can determine only the even coefficients fjk in the A more sophisticated (but technically more complicated)
expansion (2). Let us assume that plasma has axial symmetry method, suitable for measurements of EDF with an arbitrary
(in this case, k = 0, fj ≡ fj0 and Pj ≡ YJ0 are the Legendre degree of anisotropy, has been developed in [130–132]. In this
polynomials) and that the probe sheath is thin. The measured method, the measurements are performed with a flat one-sided
probe current depends on the probe orientation. Let the axis of probe having different angle positions referenced to the main
the probe have an angle c to the axis of the plasma symmetry. anisotropy direction. For the case of cylindrically symmetric
Then the second derivatives of the probe electron current is plasma, to infer the EDF from the probe measurement, the
∞
π
d 2 Ie 4π e3 al  procedure developed in [130] should be applied, whereas for
= F 2j (eV ) P2j (sin θ  sin c ) dθ  . plasma with arbitrary anisotropy, the formalism developed
dV 2 m2 j =0 0
in [131] must be applied. For all cases, the probe sheath must
(29) be thin.
 Let consider in more detail the case of the cylindrically
Here, θ is a polar angle of a normal to some element in a
auxiliary coordinate system which has a polar axis normal to symmetric plasma typical for a plasma experiment. The main
the plane containing the symmetry axis of the plasma and the formula of the method defining the coefficient fj = fj0 of the
probe, and expansion (2) is

∞  

∂ eV (2j + 1)m2 1 
F2j (eV ) = f2j (eV ) − f2j (eV ) P2j dε. fj (eV ) = (Ie (V , cos f )
eV ∂(eV ) ε 4π e3 Sp −1

∞
(30) + Ie (ε, cos f )Rj (V , ε) dε)Pj (cos f ) d(cos f ).
Thus changing c , it is possible to obtain a system of equations eV

for finding fj . However, the equations contain the coefficients (31)

fj with even j alone. Therefore, this method does not give us Here, f is the angle between the plasma symmetry axis and
complete information about the EDF. This method has been the normal direction of the probe conducting surface,
used for the measurements of the EDF in the near-cathode
j
region of low-pressure arc discharges [30, 128]. [2]
2−(j +1)  ε
j −2k−1
Experimental results on Ip measured at the axis of a low-
2
Rj (ε, V ) = akj , (32)
pressure (280 Pa) helium discharge with probes perpendicular eV k=0 eV

25
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

where
(2j − 2k)!(j − 2k)
akj = (−1)k (33)
k!(j − k)!(j − 2k)!
and

   j − 1, for odd j ;
j
= j 2 (34)
2 
 , for even j.
2
From (31), for example, we obtain for f0

1
m2
f0 = I  (V , cos f ) d(cos f ). (35)
4π e3 Sp −1 p

Thus, it is necessary to measure Ip for different f and to cal-

culate expression (35) to determine f0 and expression (31) to
determine of fj . A study of errors, arising during the EDF mea-
surements by a flat one-sided probe, was performed in [132].
Formulae (35) and (31) are valid for any degree of plasma
anisotropy. To increase the accuracy in strongly anisotropic
plasma, it is necessary to conduct measurements for a rather
large number of different angles f . The considered method
has been used, for example, for measurements of the drift
velocity as a function of electron energy in noble-gas plasma
discharges [133] and to study the momentum relaxation of the
electrons fluxes [134]. The energy dependence of the coeffi-
cients fj measured in a low-pressure (65 Pa) helium dc positive Figure 44. Coefficients fj in a helium low-pressure (0.5 Torr)
column for discharge current 0.5 A is shown in figure 44 [133]. positive column: f0 (1), f1 (2), f2 (3), f3 (4) and f4 (5) [134].
In this rather high-pressure plasma, the measured anisotropy
degree is very small, and the conventional Druyvesteyn method
is quite applicable. The accuracy of the inferred EDF depends
on the number K of the measurements performed at different
angles. Figure 45 shows the reproduction of the model function
by calculated EDF for different numbers of probe orientations
K. As expected, increasing the number of measurements at
different angles and, thus, the number of coefficients fj leads
to a more accurate EDF recovery.

4.4. Plasma electron spectroscopy

The sensitivity, dynamic range and energy resolution of current
EDF measurement techniques allow for the analysis of the
fine structure in the EDF tail corresponding to a tiny fraction
(10−3 –10−6 ) of the electron population. Such sensitivity in
the EDF measurement allows the recognition and evaluation
of some plasma-chemical and collisional processes that affect
the EDF structure. This method, called plasma electron
spectroscopy (PLES), is based on the measurement of the
EDF in the conditions where a part of the distribution function
is affected by a particular collisional process. In this case,
the EDF structure can be a source of information about
such a process. Usually, such analysis is possible when the
contribution of thermal electrons to the tail of the measured
EDF is minimal and does not mask the EDF structure caused
by the particular collisional process. Such conditions are
achieved at low electron temperatures, when EDF drops
sharply with electron energy. The idea of PLES has been
realized in experiments with afterglow plasma [21, 25, 32]
where the EEDF consists of two almost non-interactive groups
Figure 45. The polar diagram EDF for electrons calculated for
different numbers of probe orientations K: K = 3 (1); K = 5 (2);
K = 7 (3); K = 9 (4); model function (5) [21].
of electrons. Low-energy groups of thermal electrons with
Maxwellian distributions have electron temperatures at the
order of 0.1 eV. The distribution of energetic electron groups
is far from equilibrium and exhibits several peaks in the EDF
tail. The density of these electrons is from 104 to 106 times
less than that of the low-energy electrons. The origination of
the high-energy electrons is collisions of metastable excited
atoms between themselves and with electrons in the reactions
A∗ + A∗ → A+ + A + e,
A∗ + A∗ → A+2 + e, (36)
∗
A + e → A + e.

26
J. Phys. D: Appl. Phys. 44 (2011) 233001 Topical Review

Here, A∗ denotes an exited atom, and A+ and A+2 denote atomic

and molecular ions. The first two reactions in (36) represent
the Penning ionization process, whereas the third describes
the super-elastic collision of an exited atom with a low-energy
thermal electron.
Analysis of the measured EDF structure allows the
determination of information about the spectra and the
rate constants of the corresponding reactions (36). A
detailed analysis of the relationship between the measured
EEDF structure and the reaction-rate constant is provided
in [32]. A similar method has been studied in [135] for
the measurements of the reflection coefficients of electrons
from plasma boundaries and for the energy dependence of the
frequency of electron–electron collisions.
The authors of [136] appear to be the first to suggest
and use the PLES method to monitor and control the absolute Figure 46. Schematic diagram of an experimental device with a
component densities of a gas mixture. For this purpose, the cold cathode (C), an anode (A), and a cylindrical wall (W). A typical
structure of the discharge plasma is shown. The dashed line
exact knowledge of EEDF may not be as important because a indicates the cathode-sheath boundary. The negative (NG) and
calibration of the system can be provided independently. anode (AG) glows are the shaded regions near the cathode and
Recently, it was demonstrated that the plasma of a dc anode, respectively, and the Faraday dark space (FDS) is between
glow discharge near a cold cathode, where Te can be close the NG and AG [135].
to room temperature, can also be suitable for PLES and
analytical gas detection [137]. The EEDF measurement in a dc
glow discharge is considerably simpler and has dramatically
higher sensitivity than that in the afterglow plasma. For this
reason, a short (without positive column) dc discharge with a
cold cathode and a wall probe has been used experimentally
[137] for PLES analysis. Here, the EEDF was measured
with a negatively biased wall probe collecting only high-
energy electrons. The area of the wall probe is significantly
larger than that of the cylindrical Langmuir probe. The large
wall-probe surface area results in a dramatic increase in the
probe sensitivity. The distortion of the EEDF measurements
associated with the ion current effect is significantly reduced
due to the large wall probe radius. Consequently, the ion
current to the wall probe is practically independent of the probe
voltage. Figure 47. High-energy portion of d2 Ip /dV 2 (absolute value) in
neon (3 Torr), argon (0.5 Torr) and oxygen (20%)/argon (80%)
It is worth noting that, due to the use of a wall probe, the (0.5 Torr) dc discharge. The discharge currents were 10 mA, 2 mA
discharge volume (and device) size in an implementation of and 3 mA, respectively. The maxima at 16 eV and 11.5 eV are due to
PLES can be dramatically reduced because there is no need for collisions of neon and argon metastable atoms with slow electrons.
an insertion of a cylindrical Langmuir probe into the plasma. The maximum at ≈4 eV is due to electron detachment from oxygen,
This fact opens a possibility of building a micro-discharge i.e. O− + O → O2 + e [137].
operating at atmospheric gas pressure as a micro-scale gas
sensor. (20%)/argon (80%) mixture are shown in figure 47. the peaks
In [137], the discharge takes place between a plane disc- in the d2 Ip /dV 2 energy dependence for pure neon and argon
shaped molybdenum cathode (C) and anode (A), schematically at 16.6 eV and 11.5 eV are the result of collisions of slow
shown in figure 46. The plasma volume is bounded by a thermal electrons with the excited atoms (the third reaction
cylindrical stainless steel wall (W). The cathode and anode in (36)). The small peak in the oxygen argon mixture near 4 eV
are 2.5 cm in diameter. The distance between the cathode and is probably due to electron detachment from negative oxygen
anode is 1.2 cm. The wall W was used as a large wall probe. ions. The energetic electrons from the detachment have energy
The discharge plasma contains energetic primary electrons of 3.6 eV. Thus, this simple device demonstrates the possibility
accelerated in the near-cathode sheath in the direction of the of detection of the gas constituencies in the discharge plasma.
anode. These electrons ionize the gas thus producing the
plasma. They also create excited and metastable atoms, which 5. Conclusions
can in turn generate other groups of energetic electrons, such
as in reactions (36). The practical measurement of EEDF has a half-century history.
The maxima in d2 Ip /dV 2 measured with the wall probe Since the first EEDF measurements in the 1960s, significant
in the glow discharge plasma in neon, argon and oxygen progress in EDF-measurement theory and technique, involving

27
J. Phys. D: Appl. Phys. 44 (2011) 233001
a sophisticated analogue electronics and digital signal
processing, has been achieved. Many new interesting features
in the measured EEDF reflecting a variety of electron kinetic
effects have been found in the last decades in different kinds
of dc, pulsed and rf discharges.
However, the majority of EEDF measurements performed
with homemade and commercial plasma probe equipment are
of mediocre and even unacceptable quality because, in many
works involving plasma probe diagnostics (and specifically in
EEDF measurements), the basic requirements and limitation
of classical probe diagnostics are ignored. Similarly, the issue
of rf probe compensation in rf plasma diagnostics is frequently
ignored or formally addressed with the application of some rf
filter whose filtering function is inadequate. In many cases,
a lack of skill with analogue electronics and rf techniques
prevents obtaining the correct electrical measurements.
Therefore, the goal of this review is to increase awareness
of the problems pertaining to the relationship between the
actual plasma parameters and the probe experiment design.
Main sources of error in EEDF measurements, remedies to
avoid EEDF distortions and examples of positive resolutions
of the problems are presented here for different types of
gas-discharge plasmas. We also introduce the reader to
unconventional methods of electron-distribution diagnostics
in collisional, magnetized and anisotropic plasmas that are
still under development and remain a challenge for budding
scientists.
Acknowledgments
This work was supported by the DOE OFES (Contract
No DE-SC0001939), NSF (Grant No CBET-0903635) and
AFOSR.
References
[1] Stark J, Retschinky T and Schaposchnikoff A 1905 Ann.
Phys., Lpz. 323 213
[2] Langmuir I and Mott-Smith H M 1924 Gen. Electr. Rev. 27
449
Langmuir I and Mott-Smith H M 1924 Gen. Electr. Rev. 27
538
Langmuir I and Mott-Smith H M 1924 Gen. Electr. Rev. 27
616
Langmuir I and Mott-Smith H M 1924 Gen. Electr. Rev. 27
762
Langmuir I and Mott-Smith H M 1924 Gen. Electr. Rev. 27
810
[3] Mott-Smith H M and Langmuir I 1926 Phys. Rev. 28 727
[4] Druyvesteyn M J 1930 Z. Phys. 64 781
[5] Kagan Y M and Perel V I 1964 Sov. Phys.—Usp. 6 767
[6] Chen F F 1965 Plasma Diagnostic Techniques
ed R H Huddlestone and S L Leonard (New York:
Academic) p 113
[7] Medicus G 1967 Invited Paper at 8th Int. Conf. Phenomena
Ionized Gases (Vienna, Austria)
[8] Schott L 1968 Plasma Diagnostics ed W Lochte-Holtgreven
(Amsterdam/New York: North-Holland/Wiley)
p 674 chapter II
[9] Kozlov O V 1969 An Electrical Probe in a Plasma (Moscow:
Atomizdat)
28
Topical Review
[10] Swift J D and Schwar M J R 1970 Electrical Probes for
Plasma Diagnostics (London: Iliffe Books)
[11] Chung P L, Talbot L and Touryan K J 1975 Electric Probes in
Stationary and Flowing Plasma (Berlin: Springer)
[12] Ivanov Y A, Lebedev Y A and Polak L S 1981 Methods of
Contact Diagnostics in Non-Equilibrium
Plasma-Chemistry (Moscow: Nauka)
[13] Cherrington B E 1982 Plasma Chem. Plasma Process. 2 113
[14] Amemija H and Sakamoto Y 1985 J. Vac. Soc. Japan 28 177
[15] Hutchinson I H 1987 Principles of Plasma Diagnostics
(New York: Cambridge University Press)
[16] Alekseev B V and Kotelnikov V A 1988 Probe Method for
Plasma Diagnostics (Moscow: Energoatomizdat)
[17] Hershkowitz N 1989 Plasma Diagnostics vol 1 (Boston:
Academic) p 113
[18] Stangeby P C 1989 Plasma Diagnostics vol 2 (Boston:
Academic) p 157
[19] Godyak V A 1990 Plasma–Surface Interaction and
Processing of Materials (Deventer: Kluwer) p 95
[20] Matthews G F 1994 Plasma Phys. Control. Fusion 36 1595
[21] Demidov V I, Kolokolov N B and Kudryavtsev A A 1996
Probe Methods for Low-Temperature Plasma Investigation
(Moscow: Energoatomizdat)
[22] Tichy M, Kudrna P, Behnke J F, Csambal C and Klagge S
1997 J. Physique IV 7 397
[23] Amemiya H 1999 Diagnostics of Ionized Gas (Saitama: The
Institute of Physical and Chemical Research (RIKEN))
[24] Ovsyannikov A A and Zhukov M F (ed) Plasma Diagnostics
2000 (Cambridge: Cambridge International Science)
[25] Demidov V I, Ratynskaia S V and Rypdal K 2002 Rev. Sci.
Instrum. 73 3409
[26] Chen F F 2003 Mini-Course on Plasma Diagnostics (Lecture
Notes on Langmuir Probe Diagnostics) (IEEE-ICOPS
Meeting, Jeju, Korea)
[27] Pfau S and Tichy M 2008 Low Temperature Plasmas:
Fundamentals, Technologies and Techniques ed R Hippler
et al (Berlin: Wiley) p 131
[28] Kolobov V I 2006 J. Phys. D: Appl. Phys. 39 R487
[29] Golubovskii Y B, Kolobov V I and al-Havat 1988 Sov.
Phys.—Tech. Phys. 33 1046
[30] Demidov V I, Kolokolov N B, Mezentsev A P and
Mustafaev A S 1986 Sov. J. Plasma Phys. 12 866
[31] Gill P and Webb C E 1977 J. Phys. D: Appl. Phys. 10 299
[32] Kolokolov N B and Blagoev A B 1993 Phys.—Usp. 36 152
[33] Morgulis N D and Kravchenko A I 1975 Sov. Phys.—Tech.
Phys. 20 848
[34] Amemiya H 1988 J. Phys. Soc. Japan 57 887
[35] Amemiya H 1990 J. Phys. D: Appl. Phys. 23 999
[36] Popov T and Petkov I 1998 Vacuum 51 89
[37] Boedo J, Gray D, Chousal L, Conn R, Hiller B and
Finken K H 1998 Rev. Sci. Instrum. 69 2663
[38] Katsumata I and Okazaki M 1967 Japan J. Appl. Phys. Part 1
6 123
[39] Demidov V I, Koepke M E and Raitses Y 2010 Rev. Sci.
Instrum. 81 10E129
[40] Katsumata I 1996 Contrib. Plasma Phys. S 36 73
[41] Demidov V I, Ratynskaia S V and Rypdal K 1999 Rev. Sci.
Instrum. 70 4266
[42] Ratynskaia S V, Demidov V I and Rypdal K 2000 Rev. Sci.
Instrum. 71 1367
[43] Demidov V I, Finnegan S M, Koepke M E and Reynolds E W
2003 Rev. Sci. Instrum. 74 4558
[44] Demidov V I, Finnegan S M, Koepke M E and Reynolds E W
2004 Contrib. Plasma Phys. 44 689
[45] Gunn J P, Devynck P, Pascal J-Y, Adamek J, Duran I,
Hron M, Stockel J, Zacek F, Barina O and Hrach R 2002
Czech. J. Phys. 52 1107
[46] Adamek J et al and the ASDEX Upgrade Team 2010 Contrib.
Plasma Phys. 50 854
J. Phys. D: Appl. Phys. 44 (2011) 233001
[47] Shkarofsky I P, Johnston T W and Bachinski M P 1966 The
Particle Kinetics of Plasmas (London: Addison-Wesley)
[48] Schoenberg K F 1979 A study of the general plasma
characteristics of a high power multifilament ion source
PhD Thesis (Lawrence Berkeley Laboratory)
[49] Schoenberg K F 1980 Rev. Sci. Instrum. 51 1159
[50] Lieberman M A and Lichtenberg A J 2005 Principles of
Plasma Discharges and Materials Processing (New York:
Wiley)
[51] Godyak V A, Piejak R B and Alexandrovich B M 2002
Plasma Sources Sci. Technol. 11 525
[52] Volkova L M, Devyatov A M, Kralkina E A, Sedov N N and
Sherif M A 1975 Vestn. Mosk. Univ., Fiz. Astron. 16 502
[53] Waymouth J F 1966 J. Appl. Phys. 37 4492
[54] Little R G, Waymouth J F 1966 Phys. Fluids 9 801
[55] Piejak R B, Godyak V A and Alexandrovich B M 1992
Plasma Sources Sci. Technol. 1 179
[56] Kudryavtsev A A, Demidov V I, DeJoseph C A Jr, Adams S F
and Serditov K Yu 2009 Contrib. Plasma Phys. 49 373
[57] Godyak V A and Piejak R B 1990 Phys. Rev. Lett. 65 996
[58] Godyak V A, Piejak R B and Alexandrovich B M 1992 Phys.
Rev. Lett. 68 40
[59] Godyak V A and Piejak R B 1993 Appl. Phys. Lett. 63 3137
[60] VGPS Probe System www.plasmasensors.com
[61] Godyak V A and Maximov V N 1977 Vestn. Mosk. Univ. ser.
Fiz. 6 51
[62] Godyak V A and Alexandrovich B M 2005 Proc. 27th ICPIG,
vol 1, ed T Goto (Eindhoven: The Nederland) p 221
[63] Godyak V A, Nagorny V I and Lee D S 2009 GEC 2009
(Saratoga Springs, NY) (see also in [36])
[64] Godyak V A and Popov O I 1977 Sov. Phys.—Tech. Phys. 22
461
[65] Godyak V A and Oks S N 1979 Sov. Phys.—Tech. Phys. 24
784
[66] Paranjpe A P, McVittie J P and Self S A 1990 J. Appl. Phys.
67 6718
[67] Flender U, Nguyen Thi B H, Wiesemann K, Khromov N A
and Kolokolov N B 1996 Plasma Sources Sci. Technol.
5 61
[68] Godyak V A 2011 Plasma Sources Sci. Technol. 20 025004
[69] Cantin A and Gagne R R J 1974 3rd Int. Conf. on Gas
Discharges (London, UK) p 625
[70] Braithwaite N S, Benjamin N M P and Allen J E 1987
J. Phys. E: Instrum. 20 1046
[71] Nolle V L, Goodyear A, Hopgood A A, Picton P D and
Braithwaite N St J 2002 Knowl. Based Syst. 15 349
[72] Godyak V A and Piejak R B 1990 J. Appl. Phys. 68 3157
[73] Godyak V A and Piejak R B 1993 Appl. Phys. Lett. 63 3137
[74] Tsendin L D 1995 Plasma Source Sci. Technol. 4 200
[75] Kolobov V I and Godyak V A 1995 IEEE Trans. Plasma Sci.
23 503
[76] Kortshagen U, Bush C and Tsendin L D 1996 Plasma Source
Sci. Technol. 5 1
[77] Godyak V A and Khanneh A S 1986 IEEE Trans. Plasma
Sci. 14 112
[78] Godyak V A 2003 Plasma Phys. Control Fusion 45 A399
[79] Godyak V A, Piejak R B, Alexandrovich B M and
Smolyakov A I 2000 Plasma Sources Sci. Technol. 9 541
[80] Godyak V A and Alexandrovich B M 2004 Appl. Phys. Lett.
84 1468
[81] Godyak V A and Alexandrovich B M and Kolobov V I 2001
Phys. Rev. E 64 026406
[82] Godyak V A and Sternberg N 1991 Proc. 20th ICPIG vol 2
ed V Palleschi and M Vaselli (Il Ciocco, Italy) p 661
[83] Godyak V and Sternberg N 1990 Phys. Rev. A 42 2299
[84] Maresca A, Orlov K and Kortshagen U 2002 Phys. Rev. E
65 056405
[85] Rayment S W and Twiddi N D 1969 Brit. J. Appl. Phys. D
2 1747
29
Topical Review
[86] Alexandrovich B M, Godyak V A and Lister G G 2004 Proc.
of the 10th Int. Symp. on Science and Technology of Light
Sources ed G Zissis (Toulouse, France) p 283
[87] Biondi M A 1954 Phys. Rev. 93 1136
[88] Dovzhenko V A, Ershov A P and Solntsev G S 1974 Sov.
Phys.—Tech. Phys. 19 538
[89] Ershov A P, Dovzhenko V A and Solntsev G S 1977 Vestn.
Mosk. Univ., Fiz. 6 25
[90] Vasil’eva I A 1974 High Temp. 12 409
[91] Ershov A P, Dovzhenko V A, Kuzovnilov A A and Oks S N
1981 Sov. J. Plasma Phys. 7 334
[92] Godyak V A, Piejak R B and Alexandrovich B M 1993 J.
Appl. Phys. 73 3657
[93] Sudit I D and Woods R C 1994 J. Appl. Phys. 76 4488
[94] Ivanov Y A, Lebedev Y A, and Polak L S 1977 Phys. Plasma
3 146
[95] Asvadurov K D and Vasil’eva I A 1975 Sov. Phys.—Tech.
Phys. 20 996
[96] Demidov V I, Ratynskaia S V and Rypdal K 2001 Rev. Sci.
Instrum. 72 4106
[97] Klagge S 1975 Beitr. Plasmaphys. 15 309
[98] Thomas T L and Battle E L 1970 J. Appl. Phys. 41 3428
[99] Rayment S W and Twiddy N D 1973 J. Phys. D: Appl. Phys.
6 2242
[100] Godyak V A 2005 Phys. Plasmas 12 055501
[101] Godyak V A, Kasouit S and Bulkin P, to be published
[102] Roca i Cabarrocas P, Bulkin P, Daineka D, Dao T H,
Leempoel P, Descamps P, Kervyn de Meerendr T and
Charliac J 2008 Thin Solid Films 516 6834
[103] Fox-Lion N, Metzler D, Oehrlein G S and Godyak V A, to be
published
[104] Demidov V I and Kolokolov N B 1981 Sov. Phys.—Tech.
Phys. 26 533
[105] Hannemann M 2008 Contrib. Plasma Phys. 48 446
[106] Amemiya H 1976 Japan J. Appl. Phys. 15 1767
[107] Demidov V I 1983 Int. J. Electron. 54 183
[108] Demidov V I, Kolokolov N B and Toronov O G 1984 Sov.
Phys.—Tech. Phys. 29 230
[109] Blagoev A B, Demidov V I, Kolokolov N B and Toronov O G
1981 Sov. Phys.—Tech. Phys. 26 1179
[110] Volkova L M, Demidov V I, Kolokolov N B and
Kral’kina E M 1984 High Temp. 22 612
[111] Demidov V I and DeJoseph C A Jr 2005 Rev. Sci. Instrum.
76 086105
[112] Waldmann O and Fussmann G 2008 Contrib. Plasma Phys.
48 534
[113] Lobbia R B and Gallimore A D 2010 Phys. Plasmas
17 073502
[114] Swift J D 1962 Proc. Phys. Soc. Lond. 79 697
[115] Lukovnikov A I and Novgorodov M Z 1971 Brief
Communications Physics #1, 27
[116] Golubovsky Y B, Zakharova V M, Pasunkin V I and
Tsendin L D 1981 Sov. J. Plasma Phys. 7 340
[117] Popov Tsv. K, Dimitrova M, Dias F. M, Tsaneva V N,
Stelmashenko N A, Blamire M G and Barber Z H 2006
J. Phys. Conf. Ser. 44 60
[118] Popov T K, Dimitrova M and Dias F M 2004 Vacuum 76 417
[119] Gorbunov N A, Kolokolov N B and Latyshev F E 2001 Tech.
Phys. 46 391
[120] Arslanbekov R R, Khromov N A and Kudryavtsev A A 1994
Plasma Sources Sci. Technol. 3 528
[121] Malkov M A 1991 High Temp. 29 329
[122] Demidov V I, Ratynskaia S V, Rypdal K and Armstrong R J
1999 Phys. Plasmas 6 350
[123] Rypdal K, Grønvoll E, Øynes F, Fredriksen Å, Armstrong R,
Trulson J and Pécseli H L 1994 Plasma Phys. Control.
Fussion 36 1099
[124] Popov T K, Ivanova P, Stöckel J and Dejarnac R 2009
Plasma Phys. Control. Fussion 51 065014
J. Phys. D: Appl. Phys. 44 (2011) 233001
[125] Popov T K, Ivanova P, Dimitrova M, Kovačič J, Gyergyek T
and Čerček M 2011 Plasma Sources Sci. Technol. 20
at press
[126] Kagan Y M, Lavrov B P and Lyaguschenko R I 1977 Sov.
Phys.—Tech. Phys. 22 349
[127] Abroyan M A, Demidov V I, Kagan Y M, Kolokolov N B
and Lavrov B P 1975 Opt. Spectrosc. 39 12
[128] Mezentsev A P and Mustafaev A S 1985 Sov. Phys.—Tech.
Phys. 30 1319
[129] Mezentsev A P, Mustafaev A S and Fedorov V L 1998
Electron Kinetics and Applications of Glow Discharges
ed U Kortshagen and L D Tsendin (New York: Kluwer)
p 531
[130] Fedorov V L 1985 Sov. Phys.—Tech. Phys. 30 584
30
Topical Review
[131] Woods R C and Sudit I D 1994 Phys. Rev. E 50 2222
[132] Lapshin V F and Mustafaev A S 1989 Sov. Phys.—Tech.
Phys. 34 150
[133] Mezentsev A P, Mustafaev A S and Fedorov V L 1988
J. Phys. D: Appl. Phys. 21 1464
[134] Lapshin V L, Mezentsev A P and Mustafaev A S 1989
J. Phys. D: Appl.Phys. 22 857
[135] Demidov V I and Kolokolov N B 1987 Sov. Phys. J.
30 97
[136] Volkova L M, Demidov V I, Kolokolov N B and
Kralkina E A 1983 Sov. Phys.—Tech. Phys. 28 583
[137] Demidov V I, Adams S F, Blessington J, Koepke M E
and Williamson J M 2010 Contrib. Plasma Phys.
50 808